Henry A. Kautz. The logic of persistence. In Proceedings of the National Conference on Artificial Intelligence (AAAI), pages 401--405, Philadelphia, Pennsylvania, 1986. Morgan Kaufmann Publishers, Inc.  Home/Search   Document Not in Database   Summary   Related Articles   Check

....similar cases is a certain respect of causality hidden in this method: By minimizing chronologically, one tends to minimize causes rather than e#ects which is the right thing to do simply because in general causes precede e#ects. On the other hand, it has already been shown elsewhere (e.g. Kautz, 1986; Sandewall, 1993; Stein and Morgenstern, 1994 ] that the applicability of chronological minimization is intrinsically restricted to domains which do not include non deterministic information. The Tail Pipe Marauder scenario of Example 2 constitutes a simple domain which does not fall into that ....

H. Kautz. The logic of persistence. In Proc. of the AAAI, p. 401--405, Philadelphia, 1986.

....of 30 paces from him when he first realizes (at step 0) that the train will reach her in 50 time units. He begins to form a plan, and refines There is an extensive related literature, treating in turn the areas of temporal projection (e.g. Gelfond, 1988, Georgeff, 1987, Haugh, 1987, Kautz, 1986, Lifschitz, 1987a, Lifschitz, 1987b, Morris, 1988, Pearl, 1988, Shoham, 1988b, Baker, 1989 ] plan interaction (e.g. Sussman, 1973, Tate, 1975, Sacerdoti, 1975, Vere, 1983 ] and meta planning (e.g. Russell and Wefald, 1989, Brooks, 1991, Brooks, 1986, Doyle, 1988, Pollack and Ringuette, ....

....describe shortly are still first order formulas. The full specification and explanations of the formalization can be found in [Nirkhe et al. 1996 ] Our projection mechanism has commonalities with some of the chronological minimization approaches, notably [ Shoham, 1988b, Lifschitz, 1987b, Kautz, 1986 ] See [Nirkhe and Kraus, 1995 ] for detailed discussion. The i denotes the step number, and P roj List is a list of quoted formulas. We used these techniques for solving different versions of the Yale Shooting Problem [ Nirkhe and Kraus, 1995 ] 3. The CSR rule ensures that the context ....

H Kautz. The logic of persistence. In Proceedings of the 5th National Conference on Artificial Intelligence, pages 401--405. AAAI, 1986.

....to capture reasoning about effects of actions by applying circumscription and default logic but discover that their straightforward formalization leads to counterintuitive results. The claim of Hanks and McDermott led to a sequence of papers which proposed solutions to the Yale shooting problem [30, 57, 73, 72, 102, 134]. A popular group of solutions (e.g. 57, 72, 134] is based on the idea that default assumptions are applied in temporal order. Of course, autoepistemic logic does not support directly such a mode of reasoning. However, it seems that temporal order is not essential. Gelfond [30] shows how to ....

....and default logic but discover that their straightforward formalization leads to counterintuitive results. The claim of Hanks and McDermott led to a sequence of papers which proposed solutions to the Yale shooting problem [30, 57, 73, 72, 102, 134] A popular group of solutions (e.g. [57, 72, 134]) is based on the idea that default assumptions are applied in temporal order. Of course, autoepistemic logic does not support directly such a mode of reasoning. However, it seems that temporal order is not essential. Gelfond [30] shows how to formalize the Yale shooting problem in autoepistemic ....

H. Kautz. The logic of persistence. In Proceedings of the 5th National Conference on Artificial Intelligence, pages 401--405, Philadelphia, USA, August 1986. Morgan Kaufmann Publishers. -- 154 --

....an explanation of what holds at later times. 4.1. Explanations in Terms of Action Occurrences The first class of explanations we will consider are those which can be expressed entirely in terms of action occurrences, i.e. extra h propositions. We will use a version of Kautz s Stolen Car problem [23] as an illustration. Example 4.1. Let E sc = hN ; fPark; Stealg; fParkedgi, where N signifies the natural numbers, and let D sc be the domain description consisting of the following two c propositions, single h proposition and single t proposition: Park initiates Parked Steal terminates ....

Henry Kautz, The Logic of Persistence, Proceedings AAAI 86, page 401, 1986.

....to take into account the indirect changes and the facts which do not change. The problem of persistency of facts has been identi ed rst in 1969 by John MacCarthy and Patrick Hayes [16] as the frame problem. The problem of deriving indirect changes has been identi ed rst in 1986 by Henry Kautz [9] as the rami cation problem. The third fundamental problem related to actions and change is the quali cation problem and concerns the complete formulation of all preconditions necessary for the execution of an action. In this paper we will only be concerned with solutions to the two former ....

H. Kautz. The logic of persistence. In Proceedings of the 5th National Conference on Arti cial Intelligence, pages 401-405. Morgan Kaufmann Publishers, August 1986.

.... ambiguity is noted and dealt with in some way, either by changing the semantics to exclude it (e.g. by assigning a preference relation to the possible models of a given theory [ Shoham, 1988 ] or by simply acknowledging it (i.e. couching conclusions in terms of the set of possible models [ Kautz, 1986 ] In an implemented system, this ambiguity is a computational problem as well as a semantic one. In this paper, we discuss some of the sources of this ambiguity, which we treat as explicit disjunction, in the sense that ambiguous information can be interpreted as defining a set of possible ....

....consistent with the domain theory. This is the simplest form of ambiguity resulting from persistence; we discuss others in Section 5. Using both backward and forward persistence and treating their opposition in this way allows us to handle cases that using forward persistence alone does not. Kautz [ Kautz, 1986 ] describes the parking lot problem, in which an agent parks a car in a lot, leaves for a time, and then returns finding the car missing, having been stolen. Maximizing only forward persistence results in believing that the car was stolen at the last possible instant before the agent returned. ....

Kautz, Henry 1986. The logic of persistence. In Proceedings AAAI-86 Fifth National Conference on Artificial Intelligence. 401--405.

....whereas [20] uses more general time points, in the spirit of the original simplified event calculus. Second, we use properties (2) 7) of the completion of initiates and terminates, whereas [20] uses the iff completion of the definitions explicitly. 7. 3 The missing car example This problem [7] illustrates the theorem of section 6. The problem can be solved in both calculi. However, its solution requires induction in the situation calculus but not in the event calculus. The scenario is that a car is parked in a car park initially, but it is not there later. The problem is to explain ....

Kautz H., The logic of persistence, Proceedings of AAAI, page 401, 1986.

....with fewer occurring actions) and the effects of each action are given by the successor state axioms. So, in one class of axioms the shooting is done with the bullet al..igned and the other with the bullet not aligned. 4.4.6 The Stolen Car Problem. The stolen car problem was proposed by Kautz [22] to illustrate problems that arise when an incomplete set of events is provided in a reasoning task. The basic idea is that an individual parks a car in the street. Later, the person goes to get the car and finds that the car is no longer where she parked it. The intended conclusion is that the ....

Kautz, H. The logic of persistence. In Proceedings of the Fifth National Conference on Artificial Intelligence (AAAI-86) (1986), pp. 401--405.

....waiting, and ring take place in succession as in the YTS, but ring only kills the turkey if it is not hiding. The intended conclusion is that at the end of ring, the turkey is either deaf and not alive, or nondeaf and alive. Sandewall points out that this problem confutes methods like Kautz s [19] which unconditionally prefer later changes to earlier ones (and so leave the turkey unhidden and hence deaf and doomed) In an EC based approach, this variant is quite analogous to the RTS. We add an e ect axiom that Hide brings about h (eff4) and EC axioms that only Hide and Unhide can bring ....

H. Kautz. The logic of persistence. In Proc. of the 5th Nat. Conf. on AI (AAAI 86), pages 401-405, Philadelphia, PA, August 11-15 1986.

....motivated in terms other than if you do not do it, you get counterintuitive results . Causal network theory thus provides an external motivation for these choices. 10. I 0 and the others We will now see that two popular non causal approaches, namely Baker s [1] and Chronological Minimization [13, 30] can be reinterpreted as an approximation of I 0 specifically, these approaches try to generate the sufficient causes within their models, but sometimes fail to do so so once again, our preliminary work is not done well. While our comparison to chronological minimization remains completely ....

....; then for all Mb and M bt : 1. Baker s procedure selects Mb ) There exists an M bt with M bt = Mb and M bt I b 0 (T nc ) 2. M bt I b 0 (T nc ) There exists an Mb preferred by Baker s procedure with M bt = Mb . 10.2 I 0 and Chronological Minimization In chronological minimization [30, 13] (CM) one selects a model with a change of fluent value at time t only if there is no model which is the same (in terms of fluent facts) up until time t but in which no fluent changes value at time t. Again, this can be reinterpreted as approximating step 1 of I 0 . To see this, note that it is ....

H.A. Kautz. The logic of persistence. In Proceedings AAAI-86, pages 401--405, 1986.

....to temporal order: 8 Preference 1 Prediction is local For each element, P j i , we have that P j i OE P j 1 i . We also have that WD i OE P j i and L i OE WD i . Preference 1 is essentially the approach proposed in the treatments of, for example, 20, Chapter 6] 37] and that of [18]. As Stein and Morgenstern[39] and others observed, however, the above solution is not satisfactory for explanatory purposes: since change is postponed as long as possible, one will prefer the explanation occuring latest in time and, in general, there is no reason to do so. To handle this problem, ....

Henry A. Kautz. The logic of persistence. In Proceedings of the Conference of the American Association for Artificial Intelligence, pages 401--405, 1986.

....Central to reasoning about action is a satisfactory treatment of the frame problem [ Ford and Hayes, 1991 ] The nonmonotonic approaches to the frame problem can be divided into the chronological and causal. The former incorporate a temporal preference that delays change as long as possible [ Kautz, 1986; Shoham, 1988 ] whereas the latter incorporate some underlying notion of causality [ Lifschitz, 1987; Baker, 1991; Stein and Morgenstern, 1994 ] Morgenstern [ 1996 ] discusses typical problems with existing solutions: 1) some are formulated within an overly restrictive temporal ....

Kautz, Henry A. 1986. The logic of persistence. In Proceedings of the Conference of the American Association for Artificial Intelligence. 401--405.

....Fred is alive after shooting (This was noted by Fangzhen Lin (reported in [Lifschitz, 1993c] Note that in contrast to the previous example, the conclusions obtained by chronological minimization are too strong. The method of chronological minimization and the logic of persistence proposed by Kautz[1986] both fail to handle Example 4 from Section 2.4.2 correctly both conclude incorrectly that the gun was unloaded initially and that Fred died during the waiting [Baker, 1991] Many methods proposed in the literature have trouble dealing with the indirect effects of actions. For instance, to get ....

Henry Kautz. The logic of persistence. In Proc. of AAAI-86, pages 401--405, 1986.

....chronological minimization is to prefer models which delay change as much as possible. For the Yale Shooting Problem, this approach will deliver the intended model where Fred dies. Unfortunately, there is no reason to prefer change to happen later rather than sooner. The Stolen Car Scenario [Kau86] is an example which highlights the arbitrariness of chronological minimization: Say a person parks his car. He waits one hour, and then another hour. When he returns to the parking space at 2pm he nds the car missing it has been stolen The car must have been stolen either during the rst wait ....

H. Kautz. The logic of persistence. Proceedings of the AAAI 86, pages 401-405, 1986.

....: Pn (2) where A is an action expression, and each of F , P 1 , Pn (n 0) is a fluent expression. If n = 0, then we will write it as A causes F . A domain description D is a set of v propositions and e propositions. For example, the domain description D scp for the Stolen Car Problem [12] is as follows: D scp = f initially :Stolen. Steal causes Stoleng. Semantics The semantics of A is defined by using states and transitions. A state is a set of fluent names. Given a fluent name F and a state oe, we say that F holds in oe if F 2 oe; F holds in oe if F 62 oe. A transition ....

Kautz, H.A., The logic of persistence, Proc. of the AAAI86, 1986, 401--405

....at certain times or states, we want to reason forward and backward in time, inferring actions from changes of fluents and inferring changes of fluents from actions. This paper is in line with the work related to the action description language A. Now let s consider the Stolen Car Problem (SCP) [8]. Let the SCP domain have one fluent name Stolen and two action names W ait and Steal. Then, its description D 1 in A is as follows: initially :Stolen Stolen after W ait; W ait; W ait Steal causes Stolen Published in Proc. of APPIA GULP PRODE 96, Joint Conference on Declarative Programming, ....

Kautz, H.A., The logic of persistence, Proc. of the AAAI86, 1986, 401--405

....we have the final result D(Shooting 2) j= AT 2 :Alive after Shoot, from which it is concluded that action Shoot is not abnormally executed. 5.3 A further example Example 10 Car Stolen example revisited. Let us consider this classic action scenario which is related to temporal postdiction [14]. Peter parked his car in the garage. After two periods of time (e.g. action Wait was taken place twice sequentially) he found that his car was gone. The question is: during which period was the car stolen We first formalize this domain in language AT 2 . We treat Wait as an action without ....

Kautz, H., The logic of persistence. In Proceedings of AAAI-86, pp401-405. 1986.

....belief states for the following example with a STRIPS like approach [Fikes and Nilsson, 1971] for reasoning about actions. Example 7. 2 The Yale shooting problem: This well known problem for reasoning about actions from [Hanks and McDermott, 1987] has been studied by many researchers (e.g. [Kautz, 1986], Baker, 1989] It is about whether Fred is alive after someone loads a gun, waits, and then shoots at him. The commonsense beliefs of this problem can be formalized as Pre Conditions and Post conditions where Actions are enclosed in square brackets as follows: 1) Pre cond: load the gun] ....

Henry A. Kautz. The logic of persistence. In Proceedings AAAI'86, pages 401--405, Philadelphia, 1986. AAAI Press.

....usual in modal logic, 2True (t; is equivalent to 3True (t; 35 minimizing chronologically, one tends to minimize causes rather than effects which is the right thing to do simply because in general causes precede effects. On the other hand, it has already been shown elsewhere (e.g. Kautz, 1986; Sandewall, 1993; Stein and Morgenstern, 1994 ] that the applicability of chronological minimization is intrinsically restricted to domains which do not include non deterministic information. This is best illustrated with the Tail Pipe Marauder scenario of Example 2. The following formula ....

Henry Kautz. The logic of persistence. In Proceedings of the AAAI National Conference on Artificial Intelligence, pages 401--405, Philadelphia, PA, August 1986.

....problem as an example (cf. Section 1.3) In particular, they showed that an apparently natural formalization in circcumscription turned out to have consequences that were unexpectedly weak. Various solutions to the Yale Shooting problem were proposed in the subsequent literature. Some, such as [ Kautz, 1986 ] and [ Shoham, 1986 ] employed alternatives to the above mentioned systems of nonmonotonic reasoning. Others, such as [ Lifschitz, 1987 ] Gelfond, 1988 ] Morris, 1988 ] Baker, 1989 ] Apt and Bezem, 1990 ] Baker, 1991 ] and [ Lifschitz, 1991 ] discovered workable ....

Henry Kautz. The logic of persistence. In Proc. of AAAI-86, pages 401--405, 1986.

....model for A satises every instance of Circ(A; P ; Z) McCarthy [79] shows how to formalize many important forms of commonsense reasoning using circumscription. Circumscription is one of the most prominent approaches to reasoning about actions where the frame problem provides the key challenge [48, 61, 1, 2]. Circumscription has been used to capture, e.g. semantics of logic programs [60] The relationships between default logic and circumscription as well as between autoepistemic logic and circumscription have been investigated by a number of researchers [24, 25, 41, 62, 52] Etherington s [25] and ....

H. Kautz. The logic of persistence. In Proceedings of the 5th National Conference on Articial Intelligence, pages 401405, Philadelphia, USA, August 1986. Morgan Kaufmann Publishers.

No context found.

Henry A. Kautz. The logic of persistence. In Proceedings of the National Conference on Artificial Intelligence (AAAI), pages 401--405, Philadelphia, Pennsylvania, 1986. Morgan Kaufmann Publishers, Inc.

No context found.

H. Kautz. The logic of persistence. Proceedings of the AAAI 86, pages 401-405, 1986.

No context found.

Henry A. Kautz. The logic of persistence. In Proceedings of the National Conference on Artificial Intelligence (AAAI), pages 401--405, Philadelphia, Pennsylvania, 1986. Morgan Kaufmann Publishers, Inc.

No context found.

H. A. Kautz. The logic of persistence. In Proceedings of the Fifth National Conference on Artificial Intelligence, pages 401--405, 1986.

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