A. Haas. The Case for Domain-Specific Frame Axioms. In F. Brown, editor. The frame problem in artificial intelligence. Proceedings of the 19 workshop, pages 343-348, Los Altos, California.  Home/Search   Document Not in Database   Summary   Related Articles   Check

....effect of circumscription is easier to investigate. Moreover, the formulation in terms of NATs is applicable to nondeterministic actions. Kartha [ 5 ] showed that this is not the case for Baker s original solution. 5. 3 Explanation Closure Another approach to the frame problem, developed by Haas [ 4 ] , Schubert [ 22 ] 23 ] and Reiter [ 19 ] is based on the idea of explanation closure. The process of generating explanation closure axioms can be conveniently described in terms of NATs. We will illustrate this fact with an example borrowed from [ 19 ] The language has variables for ....

Andrew Haas. The case for domain-specific frame axioms. In Frank M. Brown, editor, The Frame Problem in Artificial Intelligence, Proc. 1987 Workshop, 1987.

....frame axioms are required, so the representational aspect of the Frame Problem is addressed with the alternative notion of successor state axioms just as well. The inferential advantage of the alternative design shows if we represent the Much like [13] roots in the axiomatization technique of [6], the foundations for the alternative form of successor state axioms were laid in [10] collection of fluents that are true in a situation s by equating the atomic formula Holds(f, s) with the conditions on f to hold in s. The following formula, for instance, constitutes a suitable description of ....

A. R. Haas. The case for domain-specific frame axioms. In F. M. Brown, editor, The Frame Problem in Artificial Intelligence, pages 343--348, Los Altos, CA, 1987.

....performed. Many solutions to the frame problem are based on non monotonic logics, e.g. 18, 3, 6] Also, there has been interest in the development of solutions based on standard monotonic logics. In particular, we appeal to the solution proposed by Reiter [11] which is based on the work of Haas [2], Pednault [9] and Schubert [16] In this section, we present Reiter s monotonic solution to the frame problem. A basic theory of action in the situation calculus consists of the following sets of axioms: ffl A set Sigma bd , which includes The set Sigma b of basic situation calculus ....

HAAS, A. R. The Case for Domain-Specific Frame Axioms. In The Frame Problem in Artificial Intelligence, Proceedings of the 1987.

....calculus. The situation calculus is a very appealing language which has been used to investigate many problems related to formal reasoning about change. For example, the language has been the formalism of choice for a great number of researchers interested in the so called frame problem (e.g. [40, 20, 21, 27, 54, 33, 9, 47]) So much so, that some researchers identify the frame problem as characteristic of theories based on the situation calculus, instead of as a problem inherent to the formalization of dynamic systems. Unfortunately, as has been pointed out elsewhere (e.g. 16, 23, 55] the original situation ....

....names axioms for fluents, which we describe later. 2.2 The Frame Problem. One of the hardest problems encountered in reasoning about actions has been the so called frame problem. This problem has attracted much interest in the AI community and many solutions for it have been proposed (e.g. [18, 20, 27, 58, 47, 54, 33]) Basically, the problem is understood as one of finding a way to succinctly specify the effects of actions, given the situations in which they are performed. Most solutions to the frame problem (e.g. 58, 27, 33] are based on non monotonic logics. Lately, there has been interest in the ....

[Article contains additional citation context not shown here]

Haas, A. R. The Case for Domain-Specific Frame Axioms. In The Frame Problem in Artificial Intelligence, Proceedings of the

....procedure for generating, from these effect axioms, all the frame axioms. If possible, we also want a parsimonious representation for these frame axioms (because in their simplest form, there are too many of them) 2. 4 A Simple Solution to the Frame Problem By appealing to earlier ideas of Haas [1], Schubert [14] and Pednault [9] Reiter [11] proposes a simple solution to the frame problem, which we illustrate with an example. Suppose that (1) 2) and (3) are all the effect axioms for the fluent broken, i.e. they describe all the ways that an action can change the truth value of broken. ....

A. R. Haas. The case for domain-specific frame axioms. In F. M. Brown, editor, The frame problem in artificial intelligence. Proceedings of the 1987.

....of OE f in the successor state axioms for functional fluents guarantees the Markov property: The value of a functional fluent in a successor situation is determined entirely by properties of the current situation, and not by any other situation. Following earlier ideas of Pednault [27] Haas [12], Schubert [37] and Davis [4] Reiter [31] shows how to solve the frame problem for deterministic actions. The resulting solution yields axioms with exactly the syntactic form of successor state axioms, which is why in this paper we focus on these. Basic Action Theories Henceforth, we shall ....

A. R. Haas. The case for domain-specific frame axioms. In F. M. Brown, editor, The frame problem in artificial intelligence. Proceedings of the 1987.

....kinds of axioms are used: 1. Actions imply their preconditions and effects (meaning that for an action occurrence at time t, its preconditions must hold at t and its effects hold at t 1) 2. Non parallel composable actions may not occur at the same time step; 3. Explanatory frame axioms (Haas 1987; Schubert 1989) which state that if a fluent changes its truth value between states, then one of the actions that changes it must have occurred; 4. The initial state holds at time 0 and the goals hold at a given final time point n. Following are schemas for axioms of types (1) 3) where the ....

Haas, A. 1987. The case for domain-specific frame axioms. In Brown, F., and Lawrence, K., eds., The Frame Problem in Artificial Intelligence, Proceedings of the 1987 Workshop. Morgan Kaufmann. Los Altos, CA.

....function result maps an action (pickup(o) and a state (s) to the resulting state. One of the primary motivations for the development of circumscription was to be a formal tool for specifying such frame axioms. Several researchers have observed that there is another way of writing frame axioms [Haas 1987, Schubert 1989, Pelavin 1990 (this volume, section 4.6) This is to state that if a particular property did change when any action was performed, then that action must be one of the actions known to change that property. In this example, suppose painting and burning are the only actions known to ....

Haas, Andrew R. (1987) The Case for Domain-Specific Frame Axioms, in The Frame Problem in Artificial Intelligence, Proceedings of the 1987 Workshop, Lawrence, Kansas, F. M.

....is these transaction specification axioms which solve the frame problem in the database setting. The equivalences in these axioms were motivated by Pednault s approach to the frame problem [9] and the appeal to quantification over transactions stems from the explanation closure axioms of Haas [3] and Schubert [14] See (Reiter [10, 12] for further discussion. Querying a Database Notice that in the above account of database evolution, all updates are virtual; the database is never physically changed. To query the database resulting from some sequence of transactions, it is necessary to ....

A. R. Haas. The case for domain-specific frame axioms. In F. M. Brown, editor, The frame problem in artificial intelligence. Proceedings of the 1987 workshop, pages 343--348, Los Altos, California, 1987. Morgan Kaufmann Publishers, Inc.

....up(x; t; s) is true, we can describe those properties of the world, for example the position of the agent s hand as a function of t, which must be true during the evolution of the process picking up. 3. 1 Successor State Axioms Reiter [25] building on the ideas of Pednault [21] and of Haas [9] and Schubert [29] proposes a solution to the frame problem for deterministic, nonconcurrent actions in the absence of state constraints. This provides a systematic way of obtaining so called successor state axioms from the effect axioms. We have to generalize these successor state axioms ....

A. R. Haas. The case for domain-specific frame axioms. In F. M. Brown, editor, The frame problem in artificial intelligence. Proceedings of the 1987 workshop, pages 343--348, Los Altos, California, 1987. Morgan Kaufmann Publishers, San Francisco, CA.

.... did not repair it: POSS(a; s) oe Broken(x; do(a; s) j (a = drop(x) Fragile(x; s) Broken(x; s) a 6= repair(x) These axioms incorporate a treatment of the classic frame problem [MH69] proposed by Reiter [Rei91] extending previous proposals by Pednault [Ped89] Schubert [Sch90] and Haas [Haa87]. In particular, Reiter shows how the successor state axioms above can be automatically generated from a collection of simple effect axioms describing only the changes that result from performing an action. Frame axioms need not be enumerated since they are entailments of the successor state ....

A. R. Haas. The case for domain-specific frame axioms. In F. M. Brown, editor, The Frame Problem in Artificial Intelligence: Proceedings of the 1987 workshop, pages 343--348. Morgan Kaufmann, San Mateo, California, 1987.

....(axioms, such as (2.1) describing the effects of actions) in classical logic. This is possible if it is assumed that the effect axioms describe all of the effects of the actions. Reiter builds on Pednault s proposal but achieves additional parsimony by quantifying over actions, as recommended by Haas [1987] and Schubert [1990] Elkan [1992] proposes an approach to formalizing action domains directly in classical logic. The reasoning formalized in classical logic is monotonic in the sense that acquiring new information can only lead to additional conclusions being drawn, not to earlier conclusions ....

Andrew Haas. The case for domain-specific frame axioms. In Frank M. Brown, editor, The Frame Problem in Artificial Intelligence, Proc. of the 1987 Workshop, 1987. 189

....the separate circumscriptions therefore reduce to predicate completions. While we remain within the class of formulae to which Lifschitz s theorem is applicable, this solution is effectively monotonic, and can be likened to the frame problem solution proposed by Reiter [1991] based on the work of Haas [1987] and Schubert [1990] We can now use the original, uncompleted formulae to formalise the Yale shooting scenario and, in the context of circumscription, we get the desired results. Let S be the conjunction of (Y1.1) to (Y1.3) and let D be the conjunction of (Y2.1) to (Y2.7) We have, CIRC[S ; ....

A.R.Haas, The Case for Domain-Specific Frame Axioms, Proceedings of the 1987 Workshop on the Frame Problem, pp. 343--348.

....describe what fluents will not change actions occur, as described below. Frame axioms are expressed either by classical frame axioms or by explanatory frame axioms. Classical frame axioms [McCarthy and Hayes, 1969] declare which fluents are left unchanged by each action. Explanatory frame axioms [Haas, 1987] specify the the set of actions that could have occurred in order to achieve a state change. For example, having airplane1 at time t 1 in JFK, but not having it in JFK at time t implies that airplane1 flew from some other airport to JFK at time t: at(airplane1,JFK,t) # at(airplane1,JFK,t 1)# ....

A. Haas. 1987. The case for domain-specific frame axioms. In The Frame Problem in Artificial Intelligence, Proceedings of the 1987 Workshop. Morgan Kaufman, 1987.

....Le notoire probleme de frame [12] tient au fait que le nombre de ces axiomes est de l ordre du produit du nombre d actions et du nombre de fluents. Notre approche incorpore un traitement du probleme de frame du a Reiter [15] inspire de travaux anterieurs de Pednault [14] Schubert [18] et Haas [4]) L idee de base est de reunir les axiomes d effets pertinents a un fluent donne et de supposer que ceux ci specifient toutes les facons dont le fluent pourrait changer de valeur. Une transformation syntactique permet al..ors obtenir un axiome d etat successeur pour le fluent, par exemple: P ....

Andrew R. Haas. The case for domain-specific frame axioms. Dans F.M. Brown, editeur, The Frame Problem in Artificial Intelligence: Proceedings of the 1987 Workshop, pages 343--348, Lawrence, KA, avril 1987. Morgan Kaufmann Publishing.

....are based on non monotonic logics, e.g. Shoham 1988; Lifschitz 1987; Lin and Shoham 1992) Also, there has been interest in the development of solutions based on standard monotonic 8 logics. In particular, we appeal to the solution proposed by Reiter (1991) which is based on the work of Haas (1987), Pednault (1988) and Schubert (1990) In this section, we present Reiter s monotonic solution to the frame problem. A basic theory of action in the situation calculus consists of the following sets of axioms: ffl A set S bd , which includes The set S b of basic situation calculus axioms ....

Haas, A. R. (1987). The Case for Domain-Specific Frame Axioms. In F. M. Brown (Ed.), The Frame Problem in Artificial Intelligence, Proceedings of the 1987 Workshop, Los Altos, CA, pp. 343--348. Morgan Kaufmann Publishers, Inc.

....Any correct solution, taken as a whole, will be nonmonotonic, even if the end product is expressed in a monotonic formalism. Of course this holds, presumptively, for less systematic proposals, which discuss frame axioms for some examples without specifying a general method for generating them [Haa87, Elk92]. Thus the so called monotonic approaches are conceptually similar to the more clearly nonmonotonic 4 In Chapter 3.5 we discuss in some detail their default theory for the Yale Shooting domain. approaches which employ variants of circumscription, autoepistemic logic, default logic and logic ....

Andrew Haas. The case for domain-specific frame axioms. In Frank M. Brown, editor, The Frame Problem in Artificial Intelligence, Proc. of the 1987 Workshop, 1987.

.... is the most general formalization of closed world reasoning, the majority of approaches to the frame problem have focussed on it [Lifschitz91] Lin91] Despite this emphasis on circumscription, a few attempts were made to solve the frame problem without appealing to nonmonotonic logics, notably [Haas87], Schubert89] and [Pednault89] In particular, Haas and Schubert first proposed the concept of an explanation closure axiom (or, as we have been calling them, change axioms) for specifying the frame axioms, while Pednault was the first to formulate a mechanism for automatically obtaining the ....

Haas, A. R., "The Case for Domain Specific Frame Axioms", in Brown, F. M. (ed.) , The Frame Problem in Artificial Intelligence. Proceedings of the 1987 Workshop , 343-348, Morgan Kaufmann Publishers, Inc., 1987.

....enables us to 1. INTRODUCTION 49 combine two (or more) effect axioms and to describe more specific events. The direct effect of such a resulting event is exactly the conjunction of the postconditions. Third, we provide frame axioms which are a variation of Haas domain specific frame axioms [10] and Schubert s explanation closure [20] i.e. if state formula P is true in a state and the occurring event does not satisfy event formula E, F or G, then P is true in the successor state . The combination of direct effects and this frame handling allows us to infer the propositions that ....

....:lift right) down right side (3.2) instead. This sentence makes a statement for the case that some event will not happen, i.e. if the table is on the ground on the right side and nobody lifts it there, then it will keep its position. This is a variation of Haas domain specific frame axioms [10] or Schubert s explanation closure [20] A further advantage is that we can combine frame axioms arbitrarily. For instance, suppose we had an action for holding the vase on the table, formally hold vase. From the two sentences (vase on table hold vase) vase on table (down left side :lift ....

A. Haas. The Case for Domain-Specific Frame Axioms. Proceedings of the Workshop on The Frame Problem in Artificial Intelligence, ed. F. Brown, 1987.

.... r t (A; B; T ) Clear t Gamma1 (C) oe :Clear t 1 (C) At least one action axioms guarantee that at each odd time step some action occurs, thus preventing fluents from changing freely due to non action: b; s; d 2 fA; B; C; Tg b 6= s; b 6= d; s 6= d; b 6= T Move r t (b; s; d) Explanatory (Haas 1987): If a fluent changes its value, some action affecting it must have occurred. For example, if block C becomes unclear, then either we have moved A or B on top of C: Clear t 1 (C) Clear t Gamma1 (C) oe Move r t (A; B; C) Move r t (A; T ; C) Move r t (B; A; C) Move r t (B; T ; C) Action ....

Haas, A. 1987. The case for domain-specific frame axioms. In Brown, F. M., ed., The Frame Problem in Artificial Intelligence, Proc. 1987 Workshop.

.... frame axioms as entailments of some uniform non monotonic policy, as for example in [8, 7] Another approach is Reiter [13] using successor state axioms and precondition axioms for actions in [13] which are from the combination of Pednault [11] and Haas Schubert s explanation closure axioms [6, 15]. Although the solution of [13] is very elegant and parsimonious, it only applies in a class of domains satisfying the completeness assumption: For any action a and any fluent F , all the conditions under which a can lead to F becoming true or false in the successor state are known and ....

Haas, A., The case for domain-specific frame axioms, Proc. of the Workshop on the Frame Problem in Artificial Intelligence, Morgan Kaufmann Publishers, Inc., 1987, 343--348

....for the logical properties of actions; for resource usage; and for optimization objectives. The logical properties of STRIPS operators are encoded by the following kinds of axioms: i) explanatory frame axioms (if a state change occurs, one action that could account for it must have taken place (Haas 1987; Schubert 1989; Kautz, McAllester, Selman 1996) ii) an occurring action implies its effects and preconditions; iii) exclusiveness of logically conflicting actions; and (iv) state invariant axioms in the style of (Kautz Selman 1998b) See the references cited above for the details of the ....

Haas, A. 1987. The case for domain-specific frame axioms. In Brown, F., and Lawrence, K., eds., The Frame Problem in Artificial Intelligence, Proceedings of the 1987 Workshop. Morgan Kaufmann. Los Altos, CA.

....problem [8] arises because the number of these frame axioms is of the order of the product of the number of fluents and the number of actions. Our approach incorporates a treatment of the frame problem due to Reiter [12] who extends previous proposals by Pednault [11] Schubert [14] and Haas [5]) The basic idea behind this is to collect all effects axioms about a given fluent and assume that they specify all the ways the value of the fluent may change. A syntactic transformation can then be used to obtain a successor state axiom for the fluent, for example: P oss(a; s) oe [BROKEN(x; ....

Andrew R. Haas. The case for domain-specific frame axioms. In F.M. Brown, editor, The Frame Problem in Artificial Intelligence: Proceedings of the 1987 Workshop, pages 343--348, Lawrence, KA, April 1987. Morgan Kaufmann Publishing.

No context found.

A. Haas. The Case for Domain-Specific Frame Axioms. In F. Brown, editor. The frame problem in artificial intelligence. Proceedings of the 19 workshop, pages 343-348, Los Altos, California.

No context found.

A. Haas. The Case for Domain-Specific Frame Axioms. In F. Brown, editor. The frame problem in artificial intelligence. Proceedings of the 19 workshop, pages 343-348, 19

First 50 documents  Next 50

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