Eugene Fink and Qiang Yang. Formalizing plan justifications. In Proceeding of the Ninth Conference of the Society for Computational Studies of Intelligence, pages 9--14, 1992.  Home/Search   Document Details and Download   Summary   Related Articles   Check

....in the sense that no proper subset of it hits all the deadlocks. We call such a set a minimal hitting set (MHS) and we now indicate a method to find one. 2 It will be evident that the notion of an irredundant plan is closely related to the more general concept of a well justified plan studied in [3]. Because of the simplicity of BW, we are able to work with linear plans; an indirectly redundant move is just a directly redundant one in the coresponding non linear plan with minimal ordering. Here we do not need to generate such non linear plans. For present purposes, we also do not need the ....

....n increases and finding good numerical bounds for LP(n) The most important future direction is to generalise the investigation to related phenomena in other planning domains. Although removing redundant actions is already NP complete for essentially the basic strips style class of formalisations [3], investigating tractable subclasses in a similar way as in [1] could provide useful insights into domain independant planning. ....

E. Fink & Q. Yang, Formalizing Plan Justifications, Proceedings of the Conference of the Canadian Society for Computational Studies of Intelligence (CSCI-92) (1992) 9--14.

....the gripper could be freed and reallocated at different levels where needed. The resultant plan is translated based on the resource specification of gripper in Figure 24 to produce the left plan in Figure 26. This plan is non minimal and can be post processed with known justification techniques [13] to remove redundant actions. A justified plan is one that does not contain actions which are not necessary for achieving a goal. Post processing also ensures that the translated plan is executable. This check is needed because though the planner suggested to the scheduler that resource freeing ....

....though the planner suggested to the scheduler that resource freeing and reallocating actions are available in the domain, inserting those actions may interfere with actions already present in the plan. 39 Domain translation as well as plan justification (specifically, backward justification [13]) have been fully implemented. 6.2 Action Insertion and Completeness In general, adding actions to a plan is risky because this can change its causal structure and lead to interactions. But by using domain translation in a principled manner, planning can truly tap the benefits of decoupling ....

[Article contains additional citation context not shown here]

Fink, E., and Yang, Q. Formalizing Plan Justifications. Proc. CSCSI-92, 9-14. 1992.

....translated to domainspecific actions sub plans (on left) and post processed to remove non minimal (redundant) actions (on right) specification of gripper in Figure 6.1 to produce the left plan in Figure 6.3. This plan is non minimal and can be post processed with known justification techniques [11] to remove redundant actions. A justified plan is one that does not contain actions which are not necessary for achieving a goal. Post processing also ensures that the translated plan is executable. This check is needed because though the planner suggested to the scheduler that resource freeing ....

....because though the planner suggested to the scheduler that resource freeing and reallocating actions are available in the domain, inserting those actions may interfere with actions already present in the plan. Domain translation as well as plan justification (specifically, backward justification[11]) have been fully implemented. 63 6.2 Inserted Actions and Completeness In general, adding actions to a plan is risky because this can change its causal structure and lead to interactions. But by using domain translation in a principled manner, planning can truly tap the benefits of decoupling ....

[Article contains additional citation context not shown here]

Fink, E., and Yang, Q. Formalizing Plan Justifications. Proc. CSCSI-92, 9-14. 1992. 90

....planner (Kautz Selman, 1999) which combines both the Graphplan and SATPLAN systems mentioned above. Blackbox generates a plan of optimal parallel length. The plan is then passed through a justification algorithm which minimizes its sequential length by removing sets of unnecessary actions (Fink Yang, 1992). The justified plan also includes a description of the complete state at each time step, which is easily computed by simulating execution of the plan from the initial state. Meanwhile, a type inference algorithm computes type information for all operators and objects in the domain (Fox Long, ....

Fink, E. & Yang, Q. (1992). Formalizing plan justifications.

....but also the set of actions. Such modification is already done in plan adaptation, but then only for generating a new plan from old cases, and optimizations in the sense of this article are not considered. Some preliminary studies of action set modifications appear in the literature, though. Fink and Yang (1992) study the problem of removing redundant actions from total order plans, defining a spectrum of redundancy criteria and analysing the complexity of achieving these. It is less clear that it is interesting to study action addition; adding actions to a plan could obviously not improve the execution ....

Fink, E., & Yang, Q. (1992). Formalizing plan justifications. In Proceedings of the 9th Conference of the Canadian Society for Computational Studies of Intelligence (CSCSI'92), pp. 9--14 Vancouver, BC, Canada.

....the execution order. Optimizing a plan by removing actions is possible only if it contains redundant actions, since it will otherwise not remain valid. The problem of finding such redundant actions is NP hard although some polynomial approximation algorithms have been suggested in the literature [4]. The complexity of mixed modification of sequential plans for the purpose of refitting a plan to solve another problem instance has been investigated by Nebel and Koehler [9] However, no analysis seems to exist for mixed modification with the purpose of optimizing the execution time of a plan. ....

Eugene Fink and Qiang Yang, `Formalizing plan justifications ', in Proc. 9th Conf. of the Can. Soc. Comput. Stud. Intell. (CSCSI'92), pp. 9--14, Vancouver, BC, Canada, (May 1992).

No context found.

Eugene Fink and Qiang Yang. Formalizing plan justifications. In Proceeding of the Ninth Conference of the Society for Computational Studies of Intelligence, pages 9--14, 1992.

No context found.

Eugene Fink and Qiang Yang. Formalizing plan justifications. Proceedings of the Ninth Conference of the Canadian Society for Computational Studies of Intelligence (CSCSI), pages 9--14, 1992.

No context found.

Eugene Fink and Qiang Yang. Formalizing plan justifications. In Proceedings of the Ninth Conference of the Canadian Society for Computational Studies of Intelligence, pages 9--14, 1992.

No context found.

Eugene Fink and Qiang Yang. Formalizing plan justifications. In Proceedings of the Ninth Conference of the Canadian Society for Computational Studies of Intelligence, pages 9--14, 1992.

No context found.

Eugene Fink and Qiang Yang. Formalizing plan justifications. Proceedings of the Ninth Conference of the Canadian Society for Computational Studies of Intelligence (CSCSI), pages 9--14, 1992.

....are not achieved in the resulting plan, the learner selects one of these unachieved effects as a new primary effect. 8.1. 2 Using Abstraction as a Selection Heuristic Another heuristic in choosing primary effects is based on an algorithm for generating ordered abstraction hierarchies described in [Fink and Yang, 1992b] This algorithm selects primary effects in such a way as to maximize the number of levels in an abstraction hierarchy for primary effect restricted planning. Our experience shows that using this as a heuristic is often useful for finding good selections of primary effects. The algorithm helps ....

Eugene Fink and Qiang Yang. Formalizing plan justifications. Proceedings of the Ninth Conference of the Canadian Society for Computational Studies of Intelligence (CSCSI), pages 9--14, 1992.

....plan is primary effect justified if every operator has a justified primary effect. Intuitively, primary effect justification means that no operator is used for the sake of its side effects. We presented a general discussion of justified plans in our previous research on improving the plan quality [Fink and Yang, 1992b] For example, consider the robot domain of Example 4 and suppose that the predicate robot in(x;y) is a primary effect of go(x; y) and a side effect of break(x; y) The plan (go(1,2) go(2,3) is primary effect justified for achieving the goal robot in(3) On the other hand, the plan break(1,3) ....

Eugene Fink and Qiang Yang. Formalizing plan justifications. In Proceedings of the Ninth Conference of the Canadian Society for Computational Studies of Intelligence, pages 9--14, 1992.

....by removing any subset of its operators. The task of finding such a subplan is NP complete. To satisfy the practical need for efficient planning, we present a greedy algorithm that finds a near perfect justification in polynomial time. Proofs of the italicized statements may be found in [Fink, 1992]. 2 Backward justification To formalize the notion of justified plans, we first generalize the concept of establishment defined in [Knoblock et al. 1991] to nonlinear plans. Let ff 1 and ff 2 be two operators of a correct linear plan, such that ff 1 precedes ff 2 , and l be a precondition of ff ....

....of Pi, which means that Pi is not well justified either. Thus, any well justified plan is backward justified. In other words, well justification is stronger than backward justification. It is also stronger than the justification used in ABTWEAK [Tenenberg and Yang, 1990] We found an algorithm [Fink and Yang, 1992] that computes a well justified version of a given plan in O(P Delta j Pij 4 ) time, where j Pij is the number of operator in Pi, and P is the number of preconditions of all operators in the plan, P = P ff2 Pi jPre(ff)j. 4 Perfect justification While well justified plans cannot contain ....

[Article contains additional citation context not shown here]

Eugene Fink and Qiang Yang. Formalizing plan justifications. In Proceeding of the Ninth Conference of the Society for Computational Studies of Intelligence, pages 9--14, 1992.

.... l, or ffl l is a precondition of some operator ff 1 , such that one of the effects of ff 1 is justified and no operator between ff and ff 1 achieves l Observe that the operators that do not have justified effects may be removed from the plan without violating the correctness of the plan [ Fink and Yang, 1992a ] In this paper we assume that all operators of every plan have justified effects. Definition 2 A correct plan Pi is primary effect justified if every operator of Pi has a justified primary effect. One may verify that if a planning problem has a primary effect justified solution, then the ....

Eugene Fink and Qiang Yang. Formalizing plan justifications. In Proceedings of the Ninth Conference of the Canadian Society for Computational Studies of Intelligence, pages 9--14, 1992.

No context found.

Eugene Fink and Qiang Yang. Formalizing plan justifications. In Proceedings of the Ninth Conference of the Canadian Society for Computational Studies of Intelligence (CSCSI92), pages 9--14, Vancouver, BC, 1992.

No context found.

Eugene, F., and Quiang, Y. 1992. Formalizing Plan Justifications. In Proceedings of CSCSI-92, 9-14.

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