Mark Stefik. Planning and metaplanning. In Nils J. Nilsson and Bonnie Lynn Webber, editors, Readings in Artificial Intelligence, pages 272--286. Tioga Publishing, Palo Alto, CA, 1981. 35  Home/Search   Document Not in Database   Summary   Related Articles   Check

....helpful heuristics based on simple properties of partial plans have not been forthcoming, proponents of metaplanning suggest using the full power of a problem solver to reason about which plan specialization to do next. While this idea has been discussed at length in the literature (e.g. [14], 15] it has rarely produced useful results in practice, at least in domainindependentplanning. Thereareseveralpossibleexplanationsofthis. Many discussions of metaplanning are vague and therefore are not suitable for implementation. As the use existing objects example shows, knowledge ....

STEFIK, M. Planning and metaplanning. In Readings in Artificial Intelligence, Nilsson and Webber, Eds. Tioga Publishing, Palo Alto, CA, 1981, pp. 272--286.

....[7] and DEVISER [8] Skeletal planners store successful earlier plans in a plan database. Before planning begins, goals are compared against the skeletal plans. If one or more plans in the database satisfy the current goals and the current world model, the best plan will be chosen as in MOLGEN [9]. Opportunistic planners develop plans in two stages: parts of a plan may be arranged with backtracking and later parts are linked together and enlarged as opportunities become available. Reactive systems were developed in response to several apparent drawbacks of deliberative reasoners including ....

M. Stefik, "Planning and Metaplanning", Readings in Artificial Intel., Nilsson and Weber, eds., Tioga Publishing, Palo Alto, California, 1981, pp. 272-286.

....for much longer. There is also plenty of practical evidence for importance of structured prob lems. We have already mentioned classical planning, which has considerable suc cess in tackling real world problems such as spacecraft mission planning [61, 39] and scientific experiment design [99]. Similarly, structured representations based on Bayesian networks have been used extensively in medical diagnosis applications [52, 2] An example that is closer to the stochastic systems we describe in this work is Livingston [115] a model based reactive reasoning system for diagnosing faults ....

M. J. Stefik. Planning and meta-planning. Artificial Intelligence, 16:141 169, 1981.

....the regular optimization process of the online, cost based, integrated sensing, planning, scheduling and execution EXCALIBUR agent system [2, 6] The general idea of integrated control flow planning is, of course, not new. Early work on this topic was done with the action planning system MOLGEN [10]. However, this approach took into account neither the online situation nor costs and resources. It was thus able to produce a feasible control flow for off line planning but not to optimize it with respect to certain criteria or to handle the problems of an online situation. Work on anytime ....

Stefik. M. J. 1981. Planning and Meta-Planning (MOLGEN: Part 2). Artificial Intelligence 16(2): 141--169.

....decisions arise from the choice points in non deterministic planning algorithms, and from deciding when to begin execution. Meta level control algorithms can be simple heuristics or a recursive application of the full planning algorithm. Meta level control has also been called meta level planning [40]. As this term implies, an agent can plan not only the physical actions that it will take but also the computational actions that it will take. The method for performing this planning can range from simple heuristics to recursive application of the full planner. Stefik s Molgen planner uses the ....

M. Stefik. Planning and meta-planning, 1981.

....next. For any possible plan in A, we have its name (see (3.23) Notice that, since plan generation and execution actions are plans in A, we may have a plan generation action which generates plans composed of plan generation and execution actions. This allows for a kind of meta planning [46]. As an example, the result of the action meta plan path( At(position 2 ) might be such that = plan path( 1 ,At(position 1 ) exec( 1 ) plan path( 2 ,At(position 2 ) plan best path( 3 ,At(position 2 ) if 1 6= 2 2 = 3 then . In (3.24) we extend the set of atomic ....

M. J. Stefik. Planning and Meta-Planning. Artificial Intelligence, 16:141--169, 1981.

....activated. In this respect, L is a planning language which is not only able to represent the traditional plans, but also real planning architectures as those of reactive systems and more complex embedded systems. Notice that this feature is not provided by metaplanning systems (see for example [ Stefik, 1981; Wilensky, 1979 ] since their languages are just able to represent plan formation activities. The works closest to our are those on situated planning. These systems have languages which support their capability of integrating reactivity and reasoning. Let us examine two noteworthy systems. ....

M. J. Stefik. Planning and Meta-Planning. Artificial Intelligence, 16:141--169, 1981.

....approach of a human expert in a certain field, by breaking the system into smaller subsystems. These ideas have been implemented for many problems with varying degrees of success [1, 2, 15] Implementations based on the formal theories of linear and nonlinear planning meet hard efficiency problems [4, 12, 17, 22, 25]. An efficient planner requires an intensive use of heuristic knowledge. On the other hand, a pure heuristic implementation is unique. There is no general constructive approach to such implementations. Each new problem must be carefully studied and previous experience usually can not be applied. ....

Stefik, M. (1981) Planning and meta-planning (MOLGEN: Part 2)," Artificial Intelligence, 2: 141-169.

.... problems with varying degrees of success (see, e.g. Albus, 1991; Knoblock, 1990; Mesarovich et al., 1989; Botvinnik, 1984) Implementations based on the formal theories of linear and nonlinear planning meet hard efficiency problems (McAllester and Rosenblitt, 1991; Chapman, 1987; Nilsson, 1980; Stefik, 1981; Sacerdoti, 1975) An efficient planner requires an intensive use of heuristic knowledge. Moreover, it is possible to use both dynamic and static heuristic knowledge in reducing the search variations. The dynamic knowledge can be acquired during the run time and immediately applied for search ....

Stefik, M. (1981). Planning and meta-planning (MOLGEN: Part 2),"Artificial Intelligence (pp. 141-169), 2.

....to the one step at a time process of the SMs described above: any notion of planning and problem decomposition can be turned into a PM. Some of the ones we have been considering are abstraction hierarchies [70] 71] 72] case based planning [28] 29] 37] constraint based planners [80] [81], plan reuse [20] mixedinitiative planning [65] agent planning [2] 59] 67] and others [6] 23] 84] The one we describe here is a distributed planning PM called the Horde Planner (HP) The HP tries to collect together a complete set of resources that can solve a problem (or a set of ....

Mark J. Stefik, "Planning and Meta-Planning", Artificial Intelligence Journal, Volume 16, pp. 141-169 (1981)

....makes is that agents know about the tasks, methods, and knowledge they use (and further utilize this knowledge in adapting themselves) This is essentially the TMK Autognostic view. In this sense, TMKL can be viewed as analogous to AI theories of reflection meta reasoning such as MOLGEN [Stefik, 1981], Theo [Mitchell et al. 1989] MetaAqua [Cox, 1996] etc. ffl What a designer knows about a system being designed. The claim that this perspective makes is that descriptions of tasks, methods, and knowledge facilitate redesign of systems. This view has been advanced in the MORALE project [Abowd ....

Stefik, M. (1981). Planning and meta-planning (MOLGEN: Part 2). Artificial Intelligence, 16(2).

....[ibid] each situation has its own script. Scripts can be used more than once. The use of scripts or schemata is pervasive in the field of planning and natural language understanding, both for plan recognition [Carberry, 1988; Ramshaw, 1991] and for plan synthesis [Allen, 1987; Rosenschein, 1981; Stefik, 1981b; Dean et al. 1988; Vere, 1983; Tate, 1977; Wilkins, 1984] among others) A good discussion of the sociological motivations for scripts is found in [Suchman, 1987] ffl Hierarchical decomposition. Researchers such as Sacerdoti [Sacerdoti, E.D. 1973; Sacerdoti, 1975] ABSTRIPS, NOAH) have used ....

....goals are attempted. A higher level goal completely determines the lowerlevel goal; if a lower level goal fails, replanning is performed at the higher level node. A second variety of hierarchical plans is seen in systems such as Tate s NONLIN [Tate, 1977] and Stefik s MOLGEN [Stefik, 1981a; Stefik, 1981b] Here the abstract plan is viewed as a skeletal [Friedland and Iwasaki, 1985] or partial solution to the problem; by solving the higher level, abstract goals first, the search space for possible solutions is delimited. Another difference is that replanning in these systems is allowed for ....

Mark Stefik. Planning And Meta-Planning (Molgen: Part 2). Artificial Intelligence, 16(2):141--169, May 1981.

....has been one of the most studied research topics in formal reasoning. Work has been done in mathematical logic (e.g. 15, 51, 40] in philosophical logic (e.g. 42] in logic programming (e.g. 5] in many subfields of AI, such as mathematical reasoning (e.g. 54, 11] planning (e.g. [49]) programming languages (e.g. 48] and so on. These citations are by no means exhaustive. Our interests are in theorem proving with metatheories. Similar to previous work in automated deduction, we have mechanized an object theory OT and its metatheory MT. The mechanization has been performed ....

M. J. Stefik. Planning and Meta-Planning. Artificial Intelligence, 16:141--169, 1981.

....to manipulate this knowledge effectively. Hence the need for introspective learning, and its increasing popularity across a range of AI problem solving paradigms, from planning to case based reasoning. Meta planning was an early model of introspective reasoning found in the MOLGEN planning system (Stefik, 1981). MOLGEN could, to some extent, reason about its own reasoning processes. Meta planning provided a framework for partitioning knowledge into layers, separating planning knowledge (domain knowledge and planning operators) from metaknowledge (planning strategies) Introspective reasoning is ....

Stefik, M. (1981) Planning and Meta-Planning. Artificial Intelligence, 16, pp. 141-170.

....level behaviours where planning ahead is required. MRG tactics, as shown in section 4, allow us to mix different planning techniques according to the application. The idea of reasoning about the plan formation step is known as metaplanning, see for instance [ Hayes Roth, 1985; Laird et al. 1987; Stefik, 1981; Wilensky, 1979 ] Metaplanning techniques do not provide a metalevel programming language to reason about all the planning steps (execution, monitoring, failure handling, information acquisition etc. All the basic planning activities of MRG are represented within MRG itself. This opens up the ....

M. J. Stefik. Planning and Meta-Planning. Artif. Intell., 16:141--169, 1981.

....unsolvable problems, for example, or it will reduce them to make them solvable. The integration of meta knowledge into knowledge based systems has already been discussed very early in AI [ 18 ] 4 ] 5 ] Typical meta activities considered were strategic control [ 8 ] and meta planing [ 23 ] [ 17 ] , recognition and the repair of deadlocks in problem solving [ 15 ] 18 ] 9 ] explanation of knowledge bases [ 4 ] and acting under time restrictions [ 16 ] In logic, too, meta levels have been studied (see [ 19 ] for an excellent overview) in particular wrt. self referential phrases [ ....

M. Stefik. Planning and meta-planning (molgen: Part 2). AI journal, 16:141 -- 170, 1981.

....ahead. MRG provides the ability to combine and integrate different planning techniques uniformly in one system by means of introspective planning. As far as we know, this approach is new and has never been proposed before. The idea of metaplanning is of course not new, see for instance MolGen [14] and [9, 10, 17] However, none of these systems and approaches provides a metalevel programming language to reason about all the planning activities rather than only plan generation (like plan execution and monitoring, failure handling, information acquisition and reacting to external stimuli) ....

M. J. Stefik. Planning and Meta-Planning. Artif. Intell., 16:141--169, 1981.

....AI community [31] 10] 11] Strategic control, assessing and improving one s own competence, detecting deadends in problem solving, tutoring about, or explaining a knowledge based system are typical meta activities. Although specific systems like HACKER [31] TERESIAS [11] REASON [28] MOLGEN [30] and PDP [16] were built, a ge neral framework for incorporating meta reasoning into knowledge based systems is missing. KADS introduced the strategy layer in its conceptual models in order to cope with this kind of knowledge. However, the notions at this level were so vague that it was hardly ....

M. Stefik. Planning and meta-planning (molgen: Part 2). AI journal, 16:141 -- 170, 1981.

No context found.

Mark Stefik. Planning and metaplanning. In Nils J. Nilsson and Bonnie Lynn Webber, editors, Readings in Artificial Intelligence, pages 272--286. Tioga Publishing, Palo Alto, CA, 1981. 35

No context found.

Stefik, M. 1981. Planning and metaplanning. In Nilsson, N. J., and Webber, B. L., eds., Readings in Artificial Intelligence. Palo Alto, CA: Tioga Publishing. 272--286.

No context found.

Mark Ste k. Planning and metaplanning. In Nils J. Nilsson and Bonnie Lynn Webber, editors, Readings in Arti cial Intelligence, pages 272-286. Tioga Publishing, Palo Alto, CA, 1981. 35

No context found.

Stefik M. Planning and meta-planning (MOLGEN: Part 2). Artif Intell. 1981;16(2):141-69.

No context found.

M. J. Stefik. Planning and Meta-Planning. Artificial Intelligence, 16(2):141--169, 1981.

No context found.

Stefik, M. Planning and meta-planning (MOLGEN: Part 2), Artificial Intelligence, 1981, 2: 141-169.

No context found.

M. Stefik. Planning and metaplanning (molgen: Part 2). Artificial Intelligence, 16:141--169, 1981.

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