Drummond, M., J. Bresina, and K. Swanson, "Just-In-Case Scheduling," in Proc. 12th National Conf. on Artificial Intelligence, 1994.  Home/Search   Document Details and Download   Summary   Related Articles   Check

....X3. It could then add a contingency branch to go to X2 instead, if, upon arrival at X1, there is not enough power or time available to continue to X3. Figure 8 Waypoint and utility planning for instrument placement. This contingency planning is done using an incremental Just In Case approach [5], as illustrated in Figure 9. First a seed plan is generated having maximum expected utility. That is, the plan achieves the best objectives possible given the expected resources available (time and energy) and expected consumption of those resources by the actions involved. This plan is then ....

Drummond, M., J. Bresina, and K. Swanson, "Just-In-Case Scheduling," in Proc. 12th National Conf. on Artificial Intelligence, 1994.

....schedules which do not need to be corrected during the actual execution. Examples of such approaches are redundancy based techniques such as fault tolerant real time scheduling ( GMM95, Gho96] slack based protection ( LW94] temporal protection ( CF90, Gao95] just in case scheduling ( DBS94] and others ( For classi cation and examples of scheduling under uncertainty see [DB00] The robustness of a schedule can be guaranteed only in case of small changes. Introducing e.g. temporal redundancy on a resource can make a tradeo between robustness of a schedule and degree of resource ....

M. Drummond, J. Bresina, and K. Swanson. Just-in-case scheduling. In Proc. of the Twelth National Conference on Arti cial Intelligence (AAAI-94), pages 1098-1104, Seattle, WA, 1994. AAAI Press.

....at the point of interruption is what the planner returns. Anytime planners typically select one aspect of a plan to progressively improve. Some planners (e.g. 63] improve the level of detail in a plan as time permits, initially creating only a very abstract plan. Other planners (e.g. 24] 28] [29]) progressively improve the probability of plan success. The interruptibility of these latter planners relies on being able to incrementally improve the probability of plan success. In order for such planners to function properly, it is important to model the probability of operator success as ....

....replacement is fundamentally different in that each plan in the progression is assumed to be a valid plan. The anytime progression for Elkan s algorithm is to gain increased certitude of plan validity by progressively checking negative preconditions. The Just In Case Scheduler of Drummond et al. [29] constructs a schedule, analyzes it for the parts that are most likely to fail, and adds alternatives to those points creating what they call a multiply contingent schedule. The anytime progression for this algorithm is the incremental addition of these contingencies as time permits. The ....

Mark Drummond, John Bresina, and Keith Swanson. Just-in-case scheduling. In Proceedings of AAAI-1994.

....to signify that variables are unassigned. Early results indicate that our methods allow for both easier formulation of problems and more ecient solution. Finally we will compare our algorithms with existing methods of scheduling under uncertainty for example, MDPs and Just In Case scheduling [DBS94]. ....

M. Drummond, J. Bresina, and K. Swanson. Just-in-case scheduling. In Proceedings of AAAI-94, Seattle, Washington, USA, 1994.

....of resorting to methods for finding a solution to each altered problem, it may be possible to find solutions that are less likely to fail in the first place. In this way, we may be able to forestall extra search as well as other undesirable aspects of solution failure (downtime, frustration, etc. [DBS94]) We use the term solution stability to denote this property of solutions. Within the constraint satisfaction (CSP) framework the most straightforward way to represent problems of this type is the standard dynamic constraint satisfaction (DCSP) model, in which a problem is treated as a sequence ....

M. Drummond, J. Bresina, and K. Swanson. Just-in-case scheduling. In Proceedings AAAI-94, pages 1098--1104, 1994.

....similar to the approach in Fang et al. A simple scheme would be to calculate the set of technicians, which should have met with the technician not showing up, and recursively determine the set of all affected technicians, and only reschedule their work plans. 4.3. 2 Just In Case Scheduling In [DBS, 94] Drummond, Bresina and Swanson presented an alternative approach to the Rescheduling problem called Just In Case (JIC) scheduling. Instead of rescheduling when schedule breaks, the rescheduling is done in advance on the most probable break points, taking advantage of the time available for the ....

....from using a particular scheduling method that automatically creates several distinct schedules: Just In Case scheduling. The JIC algorithm is used for supplying an alternative schedule just in case a particular event that could break the schedule should occur. The idea was proposed in [DBS, 94] which is described in chapter 4. The method requires an analysis of the problem domain in order to determine the most likely and most severe break points in a schedule. Creating several schedules. Create several similar but distinct schedules and select the most appropriate in a given ....

Drummond M., Bresina J. & Swanson K. (1994): Just-In-Case-Scheduling. Proceedings of the AAAI-94, pp. 1098 -- 1104, the AAAI press.

....and looks for a control rule (in its control module) whose LHS is true and executes its RHS. Reactive control modules were discussed in the domain of controlling mobile robots in [Fir87, GL87, Bro86] and in the papers in the collection [Mae91a] They were also discussed in [Dru86, Dru89, DB90, DBS94, Nil94, Sch95, Sch89b] for other domains. Some of the other research regarding the role of reactivity in planning and execution is described in [LHM91, Mus94, RLU90, McD90, Mit90] The main advantage of the reactive approach is that after sensing (or making observations) the agent need not ....

....formulation the prescribed action is optimal. ffl Drummond in [Dru89] says, sound SCRs guarantee that local execution choices always lead to possible goal achievement . In our formulation, local execution choices always lead to goal achievement. Drummond and his colleagues later work [DB90, DBS94] has a lot in common with our approach here. In [DB90] they present an algorithm for incremental control rule synthesis. In [DBS94] they present an algorithm for building robust schedules that takes a nominal schedule and builds contingent schedules for anticipated errors. They validate their ....

[Article contains additional citation context not shown here]

M. Drummond, J. Bresina, and K. Swanson. Just-in-case scheduling. In Proc. of the Twelth National Conference on Artificial Intelligence, pages 1098--1104, 1994.

.... of task interactions and the combinatorics of TMS models, the process of evaluating recovery options fully may require significant computational expense, e.g. trying all possible alternative ways in which a task might be achieved (O(2 n ) This is true in general contingency planning as well [5]. In contrast, in the secondary analysis algorithms, we perform more detailed, contextual, schedule analysis based on the availability of recovery options and the possibility of failure at key points. This analysis is more expensive, but, in some situations, the added expense is warranted. For ....

....domain, namely: 25 1. How can we effectively predict the performance of a schedule when there is uncertainty in the performance of methods in the schedule 2. What are the different approximations to the execution time performance measure and when is a specific approximation appropriate [5] discusses an algorithm for a specific domain namely a real telescope scheduling problem where the stochastic actions are managed by a splitting technique. Here the Just In Case scheduler pro actively manages duration uncertainty by using the contingent schedules constructed by analyzing the ....

J. Bresina, M. Drummond, and K. Swanson. Just-in-case scheduling. In Proceedings of the Twelfth National Conference on Artificial Intelligence, 1994.

....a partialorder planner called Mahinur that supports conditional planning with contingency selection. The authors concentrate on two aspects of the problem, namely, planning methods for an iterative conditional planner and a method for computing the negative impact of possible sources of failure. [2] discusses an algorithm for a specific domain namely a real telescope scheduling problem where the stochastic actions are managed by a splitting technique. Here the Just In Case scheduler pro actively manages duration uncertainty by using the contingent schedules constructed by analyzing the ....

J. Bresina, M. Drummond, and K. Swanson. Just-in-case scheduling. In Proceedings of the Twelfth National Conference on Artificial Intelligence, 1994.

....and looks for a control rule (in its control module) whose LHS is true and executes its RHS. Reactive control modules were discussed in the domain of controlling mobile robots in [Fir87,GL87,Bro86] and in the papers in the collection [Mae91] They were also discussed in [Dru86,Dru89,DB90,DBS94,Nil94,Sch95,Sch89b] for other domains. Some of the other research regarding the role reactivity in planning and execution is described in [LHM91,Mus94,RLU90,McD90,Mit90] The main advantage of the reactive approach is that after sensing (or making observations) the agent need not spend a lot ....

....our formulation the prescribed action is optimal. Drummond in [Dru89] says, sound SCRs guarantee that local execution choices always lead to possible goal achievement . In our formulation, local execution choices always lead to goal achievement. Drummond and his colleagues later work [DB90,DBS94] has a lot in common with our approach here. In [DB90] they present an algorithm for incremental control rule synthesis. In [DBS94] they present an algorithm for building robust schedules that takes a nominal schedule and builds contingent schedules for anticipated errors. They validate their ....

[Article contains additional citation context not shown here]

M. Drummond, J. Bresina, and K. Swanson. Just-in-case scheduling. In Proc. of the Twelth National Conference on Artificial Intelligence, pages 1098--1104, 1994.

....our formulation the prescribed action is optimal. Drummond in [Dru89] says, sound SCRs guarantee that local execution choices always lead to possible goal achievement . In our formulation, local execution choices always lead to goal achievement. Drummond and his colleagues later work [DB90,DBS94] has a lot in common with our approach here. In [DB90] they present an algorithm for incremental control rule synthesis. In [DBS94] they present an algorithm for building robust schedules that takes a nominal schedule and builds contingent schedules for anticipated errors. They validate their ....

....always lead to possible goal achievement . In our formulation, local execution choices always lead to goal achievement. Drummond and his colleagues later work [DB90,DBS94] has a lot in common with our approach here. In [DB90] they present an algorithm for incremental control rule synthesis. In [DBS94] they present an algorithm for building robust schedules that takes a nominal schedule and builds contingent schedules for anticipated errors. They validate their algorithm experimentally in a real telescope scheduling domain. But in these works, they do not have a formal correctness result ....

M. Drummond, J. Bresina, and K. Swanson. Just-in-case scheduling. In Proc. of the Twelth National Conference on Artificial Intelligence, pages 1098--1104, 1994.

....work to more proactive strategies. Instead of re solving the problem after a solution has been lost, we would like to avoid losing our solution in the first place. In this way, we can forestall extra search as well as other undesirable aspects of solution failure (downtime, frustration, etc. [3]) This can be done if we can find solutions that are stable in the face of change, i.e. that remain valid for the altered problem. Two criteria for evaluating performance have been suggested earlier, and these can be carried over to the present work: i) efficiency in finding a new solution, ....

M. Drummond, J. Bresina, and K. Swanson. Just-in-case scheduling. In Proceedings AAAI-94, pp. 1098--1104, 1994.

....a partial order planner called Mahinur that supports conditional planning with contingency selection. The authors concentrate on two aspects of the problem, namely, planning methods for an iterative conditional planner and a method for computing the negative impact of possible sources of failure. [2] discusses an algorithm for a specific domain namely a real telescope scheduling problem where the stochastic actions are managed by a splitting technique. Here the Just In Case scheduler pro actively manages duration uncertainty by using the contingent schedules constructed by analyzing the ....

J. Bresina, M. Drummond, and K. Swanson. Just-in-case scheduling. In Proceedings of the Twelfth National Conference on Artificial Intelligence, 1994.

.... planning) Goldman Boddy 1996) Two more recent systems that combine conditional and probabilistic planning are Weaver (Blythe Veloso 1997) and Mahinur (Onder Pollack 1997) Both these systems more closely follow the general model described above: they pro 3 The Just In Case algorithm (Drummond, Bresina, Swanson 1994) involves creating an initial schedule and building contingent schedules for the points that are most likely to fail. We focus on planning algorithms in this paper. PLAN (init, goal, T) plans f make init plan ( init, goal ) g while plan time T and plans is not empty do CHOOSE (and remove) ....

Drummond, M.; Bresina, J.; and Swanson, K. 1994. Just-in-case scheduling. In Proc. 12th Nat.

....at these points. These examples are low level issues concerning specific scheduling constraints that have received little treatment in the literature. There are also a host of higher level issues such as the robustness of schedules and the ease of rescheduling at execution time [Ow et al. 1988; Drummond et al. 1994; Hildum, 1994] relaxation of constraints in over constrained problems [Beck, 1994] and, more generally, what constitutes a good solution. For example, imagine a system that creates schedules with a probabilistic measure of the ability to execute the schedule. Rather than guaranteeing a ....

Drummond, M., Bresina, J., and Swanson, K. (1994). Just-in-case scheduling. In Proceedings of AAAI-94, pages 1098--1104, Menlo Park, CA. AAAI Press/MIT Press.

....effects when planning actions for nondeterministic environments. The penalty for this is that they must provide actions for all possible situations in order to gain completeness. Furthermore, these assumptions may lead to unintended interactions between actions. Drummond, Bresina and Swanson (Drummond et al. 1994) use a stochastic model of action duration in order to construct contingent schedules off line. In cases in which they successfully anticipate schedule breaks, they avoid the computational costs of on demand reasoning during time critical on line execution periods. Both Dean, Kaelbling, Kirman and ....

Drummond, Mark; Bresina, John; and Swanson, Keith 1994. Just-in-case scheduling. In Proceedings AAAI-94. AAAI.

....level task, if the schedule should fail during the course of execution. This type of analysis, called contingency planning can be expensive because it could involve an exhaustive search for the appropriate method that would improve schedule robustness without diminishing the criteria requirements [Bresina94]. However, the technique we describe in this paper implements an algorithm which eliminates the need to do an exhaustive search, even though it is more expensive than our non contingency scheduling approach. In this paper, we discuss contingency scheduling issues and formalize them using five ....

....similar questions namely 1. How can we effectively predict the performance of a schedule when there is uncertainty in the performance of methods in the schedule 2. What are the different approximations to the execution time performance measure and when is a specific approximation appropriate [Bresina94] discusses an algorithm for a specific domain namely a real telescope scheduling problem where the stochastic actions are managed by a splitting technique. Here the Just In Case scheduler pro actively manages duration uncertainty by using the contingent schedules built as a result of analyzing the ....

Bresina, J.;Drummond, M; Swanson, K., "Just-In-Case Scheduling", Proceedings of AAAI-94, Seattle, WA

....to schedules where appropriate. Building contingency plans is, in general, intractable, and so contingency planners tend to be slow #Draper, Hanks, and Weld, 1994; Pryor and Collins, 1996; Weld, Anderson, and Smith, 1998#. Toovercome this problem, CPS employs the Just in Case #JIC# approach #Drummond, Bresina, and Swanson, 1994#, originally developed to handle action duration uncertainty in telescope observation schedules. For the rover domain, we extended the JIC approach as follows. # To consider uncertaintyinpower consumption and data production #as well as in task duration#. # Tochoose among potential contingency ....

Drummond, M., Bresina, J., and Swanson, K. 1994. Just-in-case scheduling. Proc. of the Twelfth National ConferenceonAI.

....to schedules where appropriate. Building contingency plans is, in general, intractable, and so contingency planners tend to be slow [Draper, Hanks, and Weld, 1994; Pryor and Collins, 1996; Weld, Anderson, and Smith, 1998] To overcome this problem, CPS employs the Just in Case (JIC) approach [Drummond, Bresina, and Swanson, 1994], originally developed to handle action duration uncertainty in telescope observation schedules. For the rover domain, we extended the JIC approach as follows. ffl To consider uncertainty in power consumption and data production (as well as in task duration) ffl To choose among potential ....

Drummond, M., Bresina, J., and Swanson, K. 1994. Just-in-case scheduling. Proc. of the Twelfth National Conference on AI.

....it is there that the airmass is minimal for a star of a given declination. Of course, all other things are rarely equal, and it is often necessary to make observations well off the meridian. However, with better scheduling, it is possible to prevent this sort of limit to limit thrashing behavior (Drummond, Bresina, and Swanson, 1994). The thrashing is essentially a result of under loading the telescope with respect to the default atis group selection rules. The entire concept of underloading is only defined with respect to a particular scheduling policy, in this case, the default atis rules. It is also possible to overload a ....

....it is clearly impractical to provide any longterm loading capability for astronomers if that capability has to be manually implemented by the pa. 4. Problems solved by the apa It is possible to improve upon the default atis group selection rules by using more sophisticated scheduling techniques (Drummond, Bresina, and Swanson, 1994). Specifically, it is possible to improve the quality of observing data by more precisely scheduling groups so that observations are taken at lower airmass (on average) and so that observations are obtained at astrophysically interesting times. Additionally, for a multi user telescope, better ....

[Article contains additional citation context not shown here]

Drummond, M., Bresina, J., and Swanson, K. 1994. Just-In-Case Scheduling.

....on our problem domain and describe the current operational version of the apa system, focusing on the issues expressed above. Observation Scheduling Domain In this section, we briefly describe the observation scheduling domain. For further details about the apa architecture, see Bresina, et al. 1994), Drummond, et al. 1995) and Edgington, et al. 1996) for a characterization of the scheduling search space, see Bresina, et al. 1995) Our problem domain involves the management and scheduling of ground based, remotely located, fully automatic telescopes. With fully automatic telescopes, the ....

....jic scheduling approach, possible schedule breaks are predicted and contingent schedules are generated to recover from the most probable breaks. Since jic scheduling is not currently being employed in our operational system, it is not further discussed here; for details, see Drummond, Bresina, and Swanson (1994). APA Operations Model In this section, we describe how operations are carried out in the current apa system. Some of the apa operations are triggered according to the time of day (w.r.t. the telescope site) and some of the operations are triggered by the arrival of atis files. Since time ....

Drummond, M., Bresina, J., and Swanson, K. 1994. Just-In-Case Scheduling. In Proceedings of the Twelfth National Conference on Artificial Intelligence. Seattle, WA. AAAI Press / The MIT Press. (Available at url http://ic-www.arc.nasa.gov/ic/projects/ xfr/jic/jic.html.)

....expressed as a set of advice statements (in atis93) is built on an expectation regarding how long each group takes to execute. If this expectation is wrong, then the schedule might break during execution. We have a technique for dealing with this and it has performed well in extensive simulation (Drummond, et al. 1994), but it has yet to be tested in practice. Also, we have realized that it is rather hard to specify a load independent pointing track. The desired pointing track is actually a function of the telescope load, calculated in terms of right ascension clumping of observing targets. Our next step is ....

Drummond, M., Bresina, J., and Swanson, K. 1994. Just In Case Scheduling.

No context found.

Drummond, M., Bresina, J., and Swanson, K. 1994. Just-In-Case Scheduling. In Proc. of AAAI--94.

....so contingency planners tend to be slow. Draper et al. 1994, Pryor and Collins, 1996, Weld et al. 1998 ] To overcome this problem, PS uses Just in Case planning. The Just in Case (JIC) technique was originally developed to generate contingent observation schedules for automated telescopes [ Drummond et al. 1994 ] The basic idea is to take an existing schedule and look for the places where it is most likely to fail. The JIC scheduler then generates alternative schedules for each of those situations. The JIC scheduler starts with a sequence of tasks, where each task must be performed in a certain ....

M. Drummond, J. Bresina, and K. Swanson. Just-in-case scheduling. In Proceedings of the 12th National Conference on Artificial Intelligence, 1994.

No context found.

Drummond, M.; Bresina, J.; and Swanson, K. 1994. Just-in-case scheduling. In Proceedings of the Twelfth National Conference on Artificial Intelligence, 1098--1104. Cambridge, Mass.: AAAI.

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