Steven Minton, John Bresina, and Mark Drummond. Commitment strategies in planning: A comparative analysis. In Proceedings of the Twelfth International Joint ConferenceonArtificial Intelligence, Australia, 1991.  Home/Search   Document Not in Database   Summary   Related Articles   Check

....plan modification operators acts as the group of translations on the vectors. Topological arguments rely on the distance of the translations. Finally, an application is given in section 4; it contains steps towards a formal treatment of the comparison of the search space of the UA and TO planners [3]. It is shown in terms of the isomorphism of their respective partial plan space (when seen as a topological space) a continuous bijection over two topological spaces has been introduced as a condition for the isomorphism between these space. Note that from a mathematical point of view, this is a ....

.... plan space, refinement operator, etc) but also elaborate directions of research for classical planning issues (such as parallelism, density of solutions, specialized partial plan space movements, etc) The reader of this paper should be familiar with PPS classical planning and in particular [3]. Although the next sections are full of algebraic tools, they are all classics or so. Anyway, algebraic structures are detailed in the appendix, or else in the paper when needed. 2 A starting point Solving a (classical planning) problem is refining an initial plan until it is a solution. A plan ....

[Article contains additional citation context not shown here]

Steve Minton, John Bresina & Mark Drummond, Commitment Strategies in Planning: A Comparative Analysis, Journal of Artificial Intelligence Research 2, Morgan Kaufmann (1994), pages 227--262.

....(iii) The generalized algorithm facilitates the separation of important ideas underlying individual algorithms from brand names , and thus provides a rational basis for understanding the tradeoffs offered by various planners. We will 2 The work of Barrett and Weld [2] as well as Minton et al. [26,27] are certainly steps in the right direction. However, they do not tell the full story since the comparison there was between a specific partial order and total order planner. The comparison between different partial order planners itself is still largely unexplored. See Section 9 for a more ....

....space characterization, this means p 1) and (ii) wind up having to consider the same precondition (PTC) for establishment more than once. The examples in Figure 8 illustrate both these behaviors on a planner that only uses establishment refinement. Example 8(a) originally from Minton et al. [26]) shows that a planner without any form of book keeping may find the same solution in multiple different search branches. That is, the candidate sets of the search nodes in different branches overlap. Specifically, the ground operator sequence 030201 belongs to the candidate sets of nodes both ....

[Article contains additional citation context not shown here]

S. Minton, J. Bresina and M. Drummond. Commitment Strategies in Planning: A Comparative Analysis. In Proc. 12th IJCAI, 1991.

....conditional planner may require some explanation. In conventional, classical planning applications, nonlinear planning is usually an improvement over linear planning because fewer commitments yields a smaller search space, at a small added cost to explore each element of that search space [ Minton et al. 1991 ] How (at home) get skis) go home b) not (at home) at b) obs (road b s) clear b s) not (clear b s) go b s) Goal1 (resort s) at s) have skis) go b c) obs (road c p) clear c p) not (clear c p) go c p) at p) resort p) Goal2 Goal3 Fail o1,1 o1,2 o2,2 o2,1 ....

Minton, Steven; Bresina, John L.; and Drummond, Mark 1991. Commitment strategies in planning: A comparative analysis. In Proceedings of the 12th International Joint Conference on Artificial Intelligence. Morgan Kaufmann Publishers, Inc.

....planner may require some explanation. In conventional, classical planning applications, nonlinear planning is usually an improvement over linear planning because fewer commitments yields a smaller search space, at a relatively minimal added cost to explore each element of that search space [12]. However, it is not clear that this tradeoff operates in the same way for conditional planners. When plans have multiple branches, the savings from considering fewer orderings is likely to be much less and may not repay the cost in the added complexity of individual plan expansion actions. In ....

Minton, Steven, Bresina, John L., and Drummond, Mark, Commitment Strategies in Planning: A Comparative Analysis, Proceedings of the 12th International Joint Conference on Artificial Intelligence, Morgan Kaufmann Publishers, Inc., 1991.

....1.1 The Problems of Knowledge Sparse Planning Research into classical planning in Artificial Intelligence has for decades concentrated on theoretical issues of planning algorithms. Recent work has concentrated on, for example, the relative performance of total order vs partial order planners [3, 57, 44], the inherent computational complexity of plan generation [6, 23] extending the expressiveness of the classical 1 model [25] and general, theoretical frameworks for planning engines [31] This research has been dominated by the use of the literal or proposition as the basic level of ....

S. Minton, J. Bresina, and M. Drummond. Commitment Strategies in Planning: A Comparative Analysis. In Twelfth International Joint Conference on Artificial Intelligence, 1991.

.... precondition respectively, satisfying j = and ff; ffi P denote the positive and negative postcondition (add and delete list) respectively, satisfying ff ffi = 2 So called because it is probably the most frequent propositional variant of STRIPS, used by, for instance, Minton et al. [20] and McAllester and Rosenblitt [19] 3 Used by, for instance, Bylander [9] and Nebel and Koehler [21] 4 ffl I P denotes the initial state and G ; G Gamma P denote the positive and negative goal respectively, satisfying G G Gamma = For o = h ; j; ff; ffii O, we write ....

Steve Minton, John Bresina, and Mark Drummond. Commitment strategies in planning: A comparative analysis. In Reiter and Mylopoulos

....O(k 2 j U j) clauses, as opposed to O(k 3 j U j) in the previous two encodings. Traditional planners that use whiteknight protection strategy, such as TWEAK [Chapman, 87] have been found to be inferior to the causal link based planners because they may establish a condition multiple times [Minton et al. 91] It is thus interesting to note that the combination of white knight protection, causal link less establishment and contiguous ordering leads to a very compact causal encoding (see also Section 7) 4 Comparison with State based Encodings As mentioned earlier, state based encodings with ....

S. Minton, J. Bresina and M. Drummond, Commitment strategies in planning: a comparative analysis, Proceedings of the International Joint Conference on Artificial Intelligence (IJCAI) , 1991.

.... Apart from failing to exploit the redundancy in the causal structure, such a premature commitment may also lead to unnecessary backtracking in the planning process (especially in planners that employ goal protection) Especially troublesome in this respect are planners such as [12] 18][13] that widen the definition of protection violation to include both those nodes that intervene and delete the protected condition and those which intervene and merely reassert the protected condition. This stronger definition of un threatened and un usurped causal links is introduced to avoid ....

.... to a causal link hS ; p ; wi if v is a step other than w, 4 For the special case of single contributor causal structures, the plans produced by nonlin [21] do not in general have exhaustive validation structures, while those producedby McAllester s planner [12] and Minton et al. s UA planner [13] have exhaustive validation structures. 5 See Section 5 for a rationale for maintaining exhaustive validation structures 6 Note that by Definition 2, a step asserting p will not violate the validation hS ; p ; wi. Thus the notion of threat is stronger than that of violation. and v 62 S and ....

[Article contains additional citation context not shown here]

S. Minton, J. Bresina and M. Drummond. Commitment Strategies in Planning: A Comparative Analysis. In Proc. 12th IJCAI, 1991.

....measures, enhancing the efficiency of the solution, or non implementability of an action in a given situation) that force the reasoner to alter the sequence of actions, and or introduce delete some past actions. Under such circumstances the adaptation becomes more akin to non linear planning [Minton 91] where the goals are not totally ordered. The adaptation process should then be capable of adjusting the overall plan, by appropriate goal selection and or ordering selection . In terms of case adaptation, this becomes synonymous with selecting the next task to be carried out and finding an ....

Minton S., Bresina J. and Drummond M.: Commitment Strategies in Planning: A Comparative Analysis. Proc. IJCAI-91. Morgan Kaufmann, San Mateo, CA, 1991, pp 259 - 265.

....preconditions are known. Before we apply the notion of forbidding preconditions to generating semi ordered hierarchies, observe that this notion may be useful even in non hierarchical planning. Most planning algorithms build plans by backward chaining from the goal state (see, for example, [Minton et al. 1991]) Intermediate plans during backward chaining are incorrect, and a planning algorithm tries to achieve the correctness by inserting new operators to establish the preconditions of all operators currently in the plan, or by imposing time precedence constraints onto operators that lead to ....

....of the primary effect restriction, because it depends on a behavior of a particular planning systems. So to explain the meaning of the primary effect restriction, we briefly describe common planning algorithms. A formal overview of several main types of planning algorithms may be found in [Minton et al. 1991]. Here we present only short informal description of both linear and nonlinear planners. Linear planners. Linear planning algorithms such as GPS [Newell and H. Simon, 1972] and STRIPS [Fikes and Nilsson, 1971] build plans by backward chaining from the goal state. The preconditions of some ....

Steven Minton, John Bresina, and Mark Drummond. Commitment strategies in planning: a comparative analysis. Proc. International Joint Conference on Artificial Intelligence (IJCAI), pages 259--261, 1991.

....advantages of the Descartes approach. The final two sections review related work, and summarize some directions for further research. 2 Passive postponement There are undoubtedly many factors that can make a simple problem intractable for a given planning algorithm. As several studies have shown [6, 15, 11], POCL planners are very sensitive to the order in which decisions are made. In particular, premature commitment to planning decisions can result in state space explosions. The idea behind least commitment planning is to postpone decisions until they can only be made in one way, i.e. they are ....

S. Minton, J. L. Bresina, and M. Drummond. Commitment strategies in planning: A comparative analysis. In Proceedings of the 12th International Joint Conference on Art ificial Intelligence, pages 259--265, 1991.

....the PSN formalism first. Definition 3 An instance of the PSN planning problem is a quadruple Pi = hP; O;I; hG ; G Gamma ii where P is a finite set of atoms; 3 So called because it is probably the most frequent propositional variant of STRIPS, used by, for instance, Minton et al. [20] and McAllester and Rosenblitt [19] 4 Used by, for instance, Bylander [9] and Nebel and Koehler [21] O is a finite set of operators of the form h ; j; ff; ffii, where ; j P denote the positive and negative precondition respectively, satisfying j = and ff; ffi P denote the ....

S. Minton, J. Bresina, and M. Drummond. Commitment strategies in planning: A comparative analysis. In Reiter and Mylopoulos [22], pages 259--265.

....of the facts about a situation and the actions taken [20] Unfortunately, there are no such ideal systems yet. We are unaware of a partial order planning system which is deductive. Usually, partial order planning is formulated as a non deductive and non declarative search algorithm (see e.g. [28, 29, 22, 32, 2]) On the other hand, deductive planning systems usually deal with total order planning only (see e.g. 21, 8, 17, 25, 6] These deductive systems are often based on the situation calculus and, thus, are hampered by the technical or inferential frame problem [12, 26] To solve this problem ....

....a representative for all partial order planners. Besides extensions like disjunctions as well as conditional and universally quantified effects, which as we believe can be easily added to our resource based partial order planner, there is whole family of partial order planners (see e.g. [22, 2]) For example, in this paper we have considered only single contributor causal structures, viz. structures, where each condition in need is supported by only a single link. Multi contributor causal structures like the one discussed in [15] allow more than one link supporting a particular ....

S. Minton, J. Bresina, and M. Drummond. Commitment strategies in planning: A comparative analysis. In Proceedings of the International Joint Conference on Artificial Intelligence, pages 259--265, 1991.

....that is based on decision theory [ Dean, 1990 ] Chrisman and Simmons [ 1991 ] produce near optimal cost plans by using Markov Decision Processes to decide what to sense. Drummond and Bresina [ 1990 ] propose an algorithm that maximizes the probability of satisfying a goal. More recent work of Minton et al. 1991 ] analyzes in a rigorous way a linear and a non linear planner in terms of their overall efficiency, examining both search space complexity and time cost. Hsu [ 1990 ] proposes to plan with incomplete information by generating a most general partial plan without committing to any choice of ....

Steven Minton, John Bresina, and Mark Drummond. Commitment strategies in planning: a comparative analysis. In Proceedings of the 12th International Joint Conference on Artificial Intelligence, pages 259--265, Sydney, Australia, August 1991.

....implication of this result for someone implementing a planning system is that the most appropriate algorithm will depend on the types of problems to be solved by the planner. 1 Introduction There has been a great deal of work recently on comparing total order and partial order planning systems [1, 9], but little has been done in comparing different partial order planners themselves. Furthermore, there has been a great deal of speculation on whether a particular partial order planner is always better than another. In this paper we focus on three basic planning algorithms, snlp, nonlin and ....

Steven Minton, John Bresina, and Mark Drummond. Commitment strategies in planning: A comparative analysis. In Proceedings of the Twelfth International Joint Conference on Artificial Intelligence, Australia, 1991.

....It is difficult to conceive of a domain, however, in which the ordering suggested by the approximation is worse than a random ordering. Reasoning about causal links is easy, but they constitute a commitment by the planner. This is avoided by planners such as TWEAK [4] that make no causal links [8, 10]. The penalty paid by such planners is the increased cost and complexity of using Chapman s Modal Truth Criterion (MTC) 4] to handle interactions between subplans. Knoblock Yang [8] argued that causal link planners perform better when there are many negative threats, whereas MTC planners ....

Steven Minton, John Bresina, and Mark Drummond. Commitment strategies in planning: A comparative analysis. In Proceedings of 12th IJCAI, volume 1, pages 259--265. Morgan Kaufmann, August 1991.

....ordering constraints as necessary. There are some plans that MTC can build but that R MTC cannot (the plan of Figure 1 is one of them) the space of partial plans constructible by R MTC is simpler that the one of MTC, which can practically entail more efficiency in the construction of solutions [5]. 4. Conclusions Consistently with Chapman s remarks, we have proved that white knights are not necessary in the process of constructing plans to reach goal situations from initial situations. We have also shown that the non consideration of white knights does not prevent from finding a plan with ....

....as a verification procedure, runs in polynomial time, we can, starting from (P, P ) derive in polynomial time a plan (P, P ) with P a minimal set of ordering constraints. In many cases, this polynomial time complexity will be negligible compared to the search time for plan generation [5, 10]. Moreover, efficient algorithms are indeed available to remove unnecessary ordering constraints from totally or partially ordered plans [3, 8, 9, 10] One should note that although Chapman s implementation of Tweak never uses [the white knight] 1, page 360] it was a correct and complete ....

[Article contains additional citation context not shown here]

Minton, S., Bresina, J. and Drummond, M., Commitment strategies in planning: a comparative analysis, in:<F1.377e+04> Proceedings of 11<F1.377e+04> th<F1.377e+04><F1.409e+04> IJCAI, Sydney, Australia (1991) 259-265.

No context found.

Steven Minton, John Bresina, and Mark Drummond. Commitment strategies in planning: A comparative analysis. In Proceedings of the Twelfth International Joint ConferenceonArtificial Intelligence, Australia, 1991.

No context found.

Steven Minton, John Bresina, and Mark Drummond. Commitment strategies in planning: a comparative analysis. Proc. International Joint Conference on Artificial Intelligence (IJCAI), pages 259--261, 1991.

No context found.

Steven Minton, John Bresina, and Mark Drummond. Commitment strategies in planning: a comparative analysis. In Proceedings of the International Joint Conference on Artificial Intelligence, pages 259--265, 1991.

No context found.

S. Minton, J. Bresina, and M. Drummond. Commitment strategies in planning: A comparative analysis. In Proceedings of 12th IJCAI, 1991.

No context found.

S. Minton, J. Bresina and M. Drummond. Commitment Strategies in Planning: A Comparative Analysis. In Proc. 12th IJCAI, 1991.

No context found.

S. Minton, J. Bresina, and M. Drummond. Commitment strategies in planning: A comparative analysis. In Proceedings of the Twelfth International Joint Conference on Artificial Intelligence, pages 259--265, August 1991.

No context found.

S. Minton, J. Bresina, and M. Drummond. Commitment strategies in planning: A comparative analysis. In Proceedings of IJCAI-91, 1991.

No context found.

Steven Minton, John Bresina, and Mark Drummond. Commitment strategies in planning: a comparative analysis. Proceedings of the International Joint Conference on Artificial Intelligence, pages 259--261, 1991.

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