K. S. Decker and V. R. Lesser. Quantitative modeling of complex computational task environments. In Proceedings of the 11th National Conference on AI, 1993.  Home/Search   Document Details and Download   Summary   Related Articles   Check

....role allocations. We presented results from two distinct domains to illustrate our methodology. While the work used team oriented programming [12, 13, 14] as an example BDI approach, it is relevant to other similar techniques of modeling and tasking collectives of agents, such as Lesser et al. s [4] TAEMS approach. In other related work, role allocation based on matching of capabilities [15] and combinatorial auctions [7] has been proposed earlier. The key difference with prior work is our use of stochastic models (RMTDPs) to evaluate allocations: this enables us to compute the benefits of ....

K. Decker and V. Lesser. Quantitative modeling of complex computational task environments. In AAAI, 1993.

....role allocations. We presented results from two distinct domains to illustrate our methodology. While the work used team oriented programming [12, 13, 14] as an example BDI approach, it is relevant to other similar techniques of modeling and tasking collectives of agents, such as Lesser et al. s [4] TAEMS approach. In other related work, role allocation based on matching of capabilities [15] and combinatorial auctions [7] has been proposed earlier. The key difference with prior work is our use of stochastic models (RMTDPs) to evaluate allocations: this enables us to compute the benefits of ....

K. Decker and V. Lesser. Quantitative modeling of complex computational task environments. In AAAI, 1993.

.... has been increasingly accepted within MAS community, and mathematical and computational methods have been progressively used to develop a better understanding of the fundamental principles of organizing MAS [36] Among others, TAEMS (Task Analysis, Environment Modeling and Simulation) [11, 10] is a modeling framework for representing coordination problems in a formal, domain independent way. A TAEMS model can be used for both the analysis and simulation of coordination algorithms, and also to design organizational structures. Although, TAEMS can be extremely useful for detailed design ....

K. S. Decker and V. R. Lesser. Quantitative modeling of complex computational task environments. In Proc. of the 11th National Conf. on Artificial Intelligence, pages 217--224, 1993.

.... has been increasingly accepted within MAS community, and mathematical and computational methods have been progressively used to develop a better understanding of the fundamental principles of organizing MAS [35] Among others, TAEMS (Task Analysis, Environment Modeling and Simulation) [11, 10] is a modeling framework for representing coordination problems in a formal, domain independent way. A TAEMS model can be used for both the analysis and simulation of coordination algorithms, and also to design organizational structures. Although, TAEMS can be extremely useful for detailed design ....

K. S. Decker and V. R. Lesser. Quantitative modeling of complex computational task environments. In Proc. of the 11th National Conf. on Artificial Intelligence, pages 217--224, 1993.

....interactions. Scheduling with uncertain information requires additions to the scheduling algorithm and the monitoring of method performance to allow dynamic reaction to unexpected situations. In design to time we frame the input problem in a generic representation of task structures known as TMS [Decker and Lesser, 1993, Decker, 1995] In TMS we can describe problems in terms of how quality is incrementally accumulated over time, what soft and hard interactions there are between tasks, and how much uncertainty there is in method quality and duration. Figure 1.2 shows an example of a simple TMS task structure, ....

....algorithms in that it represents interacting methods that can be anytime algorithms or more common algorithms that only achieve quality upon completion. In this dissertation we have developed a framework for representing multiple method problems based on the TMS task structure framework [Decker and Lesser, 1993] . This framework can represent hard and soft interactions between tasks that affect the quality they achieve and duration it takes them to complete, uncertainty in task quality and duration, and in the presence and power of interactions between tasks, and higher level tasks that 11 accumulate ....

[Article contains additional citation context not shown here]

Keith S. Decker and Victor R. Lesser. Quantitative modeling of complex computational task environments. In Proceedings of the Eleventh National Conference on Artificial Intelligence, pages 217--224, Washington, July 1993.

....in computationally intensive AI environments. In DECAF, each capability is represented as a complete task reduction tree #HTN #Erol, Hendler, Nau 1994##, similar to that in RETSINA #Williamson, Decker, Sycara 1996#, with annotations drawn from the T#MS task structure description language #Decker Lesser 1993; Wagner, Garvey, Lesser 1997#. The leaves of the tree represent basic agent actions #HTN primitive actions#. The main thrust of current research is the soft realtime component of the framework. This work is leveraging the work already started using Design to Criteria #DTC# scheduling from the ....

Decker, K. S., and Lesser, V. R. 1993. Quantitative modeling of complex computational task environments. In Proceedings of the Eleventh National ConferenceonArti #cial Intelligence, 217#224.

....central idea is to endow agents with the capability to learn to choose appropriate, situation speci c coordination strategies from a set of available strategies. This idea has been implemented in a system called COLLAGE which is an extension of the GPGP (e.g. DL92] and T MS (e.g. Dec95, Dec96, DL93] frameworks. The primary di erence between the LODES COLLAGE approaches and the JPJL algorithm is that the former are very knowledge intensive whereas the latter is not. In particular, in the case of LODES the agents are required to a priori possess deep domain knowledge and in the case of ....

K.S. Decker and V.R. Lesser. Quantitative modeling of complex computational task task environments. In Proceedings of the Eleventh National Conference on Articial Intelligence, pages 217-224, 1993.

.... for each gene can be done independently # biologists wish to both make their findings widely available, yet retain control over the data # new types of analysis and sources of data are appearing constantly We have used DECAF, a multi agent system toolkit based on RETSINA [32, 12, 8] and TAEMS [11, 34], to construct a prototype multi agent system for automated annotation and database storage of sequencing data for herpesviruses [7] The resulting system eliminates tedious and always out of date hand analyses, makes the data and annotations available for other researchers (or agent systems) and ....

.... together two pieces of work: Williamson s work on information flow representations used in RETSINA [37, 36] and Decker s work on representations of how local and non local action executions effect those characteristics over which an agent expresses preferences (via a utility function) used in TMS [11, 34]. 3.2.1 RETSINA Information Flow The unique contribution of the RETSINA information flow representation used in DECAF is the declarative description of the information requirements of actions and the information producing abilities of actions [36] This is in addition to the traditional ....

Keith S. Decker and Victor R. Lesser. Quantitative modeling of complex computational task environments. In Proceedings of the Eleventh National Conference on Artificial Intelligence, pages 217--224, Washington, July 1993.

....and data distributed across a set of agents in distributed problem solving. Selectively sharing relevant partial results of a local search or data can enhance the global problem solving quality and efficiency without paying a large price for communication. In some recent studies, Decker and Lesser[12] further quantify these observations and discuss various trade offs involved in communication vs enhanced problem solving efficiency. However, much of the past work in DAI has primarily focussed on exploiting interactions among subgoals to resolve control uncertainty. This paper deals with ....

K. Decker and V. R. Lesser, " Quantitative Modeling of Complex Computational Task Environments ", in Proceedings of the Eleventh National Conference on Artificial Intelligence, pp 217-224, Washington, 1993.

....for each gene can be done independently; biologists wish to both make their findings widely available, yet retain control over the data; new types of analysis and sources of data are appearing constantly. We have used DECAF, a multi agent system toolkit based on RETSINA [21, 10, 7] and TAEMS [9, 23], to construct a prototype multi agent system for automated annotation and database storage of sequencing data for herpesviruses [6] The resulting system eliminates tedious and always out of date hand analyses, makes the data and annotations available for other researchers (or agent systems) and ....

K. S. Decker and V. R. Lesser. Quantitative modeling of complex computational task environments. In Proc. 11th AAAI, pages 217--224, 1993.

....(collection of agent actions) These building blocks are then chained together by the DECAF planner. This paradigm differs from most of the well known agent toolkits, which instead use the API approach to agent construction (e.g. 17] Functionally, DECAF is based on RETSINA [24] and taems [5]. To augment the functions of the agent Planner, the control or programming of DECAF agents can be provided via a Agent Initialization DECAF Task and Control Structures Plan File Incoming KQML messages Domain Facts and Beliefs KQML Messages Outgoing Action Modules Action Queue Pending ....

.... pieces of work: Williamson s work on information flow representations resulting in RETSINA [30, 29] and Decker s work on representations of how local and non local action executions effect those characteristics over which an agent expresses preferences (via a utility function) resulting in taems [5, 26]. 1.3.1 RETSINA Information Flow The unique contribution of the RETSINA information flow representation used in DECAF is the declarative description of the information requirements of actions and the information producing abilities of actions [29] This is in addition to the traditional ....

K. S. Decker and V. R. Lesser. Quantitative modeling of complex computational task environments. In Proceedings of the Eleventh National Conference on Artificial Intelligence, pages 217--224, Washington, July 1993.

....for each gene can be done independently; biologists wish to both make their findings widely available, yet retain control over the data; new types of analysis and sources of data are appearing constantly. We have used DECAF, a multi agent system toolkit based on RETSINA [26, 11, 8] and TAEMS [10, 28], to construct a prototype multi agent system for automated annotation and database storage of sequencing data for herpesviruses [7] The resulting system eliminates tedious and always out of date hand analyses, makes the data and annotations available for other researchers (or agent systems) and ....

Keith S. Decker and Victor R. Lesser. Quantitative modeling of complex computational task environments. In Proceedings of the Eleventh National Conference on Artificial Intelligence, pages 217--224, Washington, July 1993.

....desirable properties such as component reuse (eventually, automated via the planner) and some design time error checking. The chaining of activities can involve traditional looping and if then else constructs. This part of DECAF is an extension of the RETSINA and TAEMS task structure frameworks [13, 2]. The goals of the architecture are to develop a modular platform suitable for our research activities, allow for rapid development of third party domain agents, and provide a means to quickly develop complete multiagent solutions using combinations of domain specific agents and standard ....

Keith S. Decker and Victor R. Lesser. Quantitative modeling of complex computational task environments. In Proceedings of the Eleventh National Conference on Artificial Intelligence, pages 217--224, Washington, July 1993.

....with desirable properties such as component reuse (eventually, automated via the planner) and some design time error checking. The chaining of activities can involve traditional looping and if thenelse constructs. This part of DECAF is an extension of the RETSINA and TMS task structure frameworks [16, 2]. The DECAF Plan Editor attaches to each action a performance profile which is then used and updated internally by DECAF to provide real time local scheduling services. The reuse of common agent behaviors is thus increased because the execution of these behaviors does not depend only on the ....

....this instantiation of the architecture is capable of performing. A plan file is an ASCII representation of a Hierarchical Task Network (HTN) that details the actions and sequences to complete a task. The actual syntax of the plan file is an extension of the RETSINA and TMS structure detailed in [16, 2]. In broad terms, a plan file is tree 1 . The plan defines execution paths along the various branches of the tree and the critical measurement of complexity of the plan is the number of actions (represented as tree leaves) to be executed. Testing Scalability is a matter of observing results ....

K. S. Decker and V. R. Lesser. Quantitative modeling of complex computational task environments. In Proceedings of the Eleventh National Conference on Artificial Intelligence, pages 217--224, Washington, July 1993.

....in computationally intensive AI environments. In DECAF, each capability is represented as a complete task reduction tree (HTN (Erol, Hendler, Nau 1994) similar to that in RETSINA (Williamson, Decker, Sycara 1996) with annotations drawn from the TAEMS task structure description language (Decker Lesser 1993; Wagner, Garvey, Lesser 1997) The leaves of the tree represent basic agent actions (HTN primitive actions) The main thrust of current research is the soft realtime component of the framework. This work is leveraging the work already started using Design to Criteria (DTC) scheduling from the ....

Decker, K. S., and Lesser, V. R. 1993. Quantitative modeling of complex computational task environments. In Proceedings of the Eleventh National Conference on Artificial Intelligence, 217--224.

....with desirable properties such as component reuse (eventually, automated via the planner) and some design time error checking. The chaining of activities can involve traditional looping and if then else constructs. This part of DECAF is an extension of the RETSINA and TMS task structure frameworks [10, 2]. The goals of the architecture are to develop a modular platform suitable for our research activities, allow for rapid development of third party domain agents, and provide a means to quickly develop complete multi agent solutions using combinations of domain specific agents and standard ....

....this instantiation of the architecture is capable of performing. A plan file is an ASCII representation of a Hierarchical Task Network (HTN) that details the actions and sequences to complete a task. The actual syntax of the plan file is an extension of the RETSINA and TMS structure detailed in [10, 2]. In broad terms, a plan file is tree 2 . The plan defines execution paths along the various branches of the tree and the critical measurement of complexity of the plan is the number of actions (represented as tree leaves) to be executed. The test scenario will be a set of plan files and a set ....

Keith S. Decker and Victor R. Lesser. Quantitative modeling of complex computational task environments. In Proceedings of the Eleventh National Conference on Artificial Intelligence, pages 217--224, Washington, July 1993.

....subtask (collection of agent actions) These building blocks are then chained together by the DECAF planner. This paradigm differs from most of the well known agent toolkits, which instead use the API approach to agent construction (e.g. 8] Functionally, DECAF is based on RETSINA [3] and TAEMS[2, 13]. Although a traditional HTN planning component is in development, currently the control or programming of DECAF agents is provided via a picture based GUI called the Plan Editor. The Plan Editor can also be used to construct shared task network libraries for common multi agent protocols. The GUI ....

Keith S. Decker and Victor R. Lesser. Quantitative modeling of complex computational task environments. In Proc. 11th AAAI, pages 217--224, Washington, July 1993.

No context found.

Decker, K. and Lesser, V. "Quantitative Modeling of Complex Computational Task Environments, " Proceedings of the Eleventh National Conference on Artificial Intelligence, 1993. Forthcoming.

....of the well known agent toolkits, which instead use the API approach to agent construction #e.g. #Chauhan 1997; Boloni 1996; Petrie 1996##. Functionally, DECAF is based on RETSINA #Sycara et al. 1996; Decker et al. 1997; Decker Sycara 1997; Williamson, Decker, Sycara 1996b; 1996a# and t#ms #Decker Lesser 1993; Wagner, Garvey, Lesser 1997#. In the absence of a planning component, the control or programming of DECAF agents is provided via a GUI called the Plan Editor. The Plan Editor can also be used to construct shared task network libraries and of course to speed up planning by providing hand coded ....

.... on information #ow representations resulting in RETSINA #Williamson, Decker, Sycara 1996b; 1996a#, and Decker s work on representations of how local and non local action executions e#ect those characteristics over which an agent expresses preferences #via a utility function# resulting in t#ms #Decker Lesser 1993; Wagner, Garvey, Lesser 1997#. RETSINA Information Flow The unique contribution of the RETSINA information #ow representation used in DECAF is the declarative description of the information requirements of actions and the information producing abilities of actions #Williamson, Decker, ....

Decker, K. S., and Lesser, V. R. 1993. Quantitative modeling of complex computational task environments. In Proceedings of the Eleventh National ConferenceonArti #cial Intelligence, 217#224.

....with desirable properties such as component reuse (eventually, automated via the planner) and some design time error checking. The chaining of activities can involve traditional looping and if thenelse constructs. This part of DECAF is an extension of the RETSINA and TMS task structure frameworks [18, 2]. The DECAF Plan Editor attaches to each action a performance profile which is then used and updated internally by DECAF to provide real time local scheduling services. The reuse of common agent behaviors is thus increased because the execution of these behaviors does not depend only on the ....

K. S. Decker and V. R. Lesser. Quantitative modeling of complex computational task environments. In Proc. of the Eleventh National Conference on Artificial Intelligence, pages 217--224, Washington, July 1993.

.... is on taking interactions among subproblems into account when building solution plans, both hard interactions that must be heeded to find correct solutions (e.g. hard precedence constraints) and soft interactions that can improve (or hinder) performance (e.g. facilitates constraints[Decker and Lesser, 1993]) Previous workon design to time scheduling has examined scheduling in an actual application[Garvey and Lesser, 1993] in a simulation of an application with particularly circumscribed interactions among tasks[Garvey et al. 1993] and in a distributed simulation[Garvey et al. 1994] In the ....

....The heuristic scheduling in this work was done using an early version of the basic algorithm described in this paper. An example of a problem to be solved by the design to time scheduling algorithm is given in Figure 1. This representation of a task structure is based on the TMS modeling framework[Decker and Lesser, 1993] . In a TMS task structure the leaves of the graph represent executable computations (known as methods) and the nonleaf nodes represent tasks that achieve quality as a function of the quality of their subtasks. Each separate graph is known as a task group and represents a single independent ....

[Article contains additional citation context not shown here]

Keith S. Decker and Victor R. Lesser. Quantitative modeling of complex computational task environments. In Proceedings of the Eleventh National Conference on Artificial Intelligence, pages 217--224, Washington, July 1993.

....task relationship that is not handled by any of the existing coordination mechanisms, namely, that the hospital patient is a mutually exclusive resource. To address this new relationship, first we must define it formally as a resource constraint in our task structure representation language (taems [9, 32]) and then we define a new GPGP coordination mechanism to handle the new task relationship. We show how the existing practices can be captured using GPGP and removing some of the mechanisms. GPGP with our new resource constraint mechanism outperforms existing practice (as described by Ow et al. ....

.... inhibits (a negative variation of the soft facilitates relationship where the performance of some test within some timeframe invalidates the results of another) These task relationships indicate when the execution of one task changes the characteristics (here, primarily duration) of another task [9]. From this view of task structure, the hospital scheduling problem has these peculiar features: ffl Tasks have no redundancy. Each test can only be done by a single ancillary department. ffl The quality accumulation functions of non executable tasks are always min. This is because here all the ....

[Article contains additional citation context not shown here]

K. S. Decker and V. R. Lesser. Quantitative modeling of complex computational task environments. In Proceedings of the Eleventh National Conference on Artificial Intelligence, pages 217--224, Washington, July 1993.

....case information. In the event of a failure, a non specialist would still be able to retrieve useful plans for a task inside its area of expertise, but outside of its specialty. 4. 1 The Portfolio Monitoring Task We can represent the plans that are retrieved using taems task structures [2, 1]. taems task structures are based on abstraction hierarchies, where task plans are elaborated via a subtask relationship into acyclic directed graphs that have actions, called executable methods, at their leaves. Such structures are compatible with most planning representations, and provide the ....

Keith S. Decker and Victor R. Lesser. Quantitative modeling of complex computational task environments. In Proceedings of the Eleventh National Conference on Artificial Intelligence, pages 217--224, Washington, July 1993.

....with desirable properties such as component reuse (eventually, automated via the planner) and some designtime error checking. The chaining of activities can involve traditional looping and if then else constructs. This part of DECAF is an extension of the RETSINA and TMS task structure frameworks [11, 2]. Unlike traditional software engineering,the DECAF PlanEditor can attach to each action a performance profile which is then used and updated internally by DECAF to provide real time local scheduling services. The reuse of common agent behaviors is thus increased because the execution of these ....

....this instantiation of the architecture is capable of performing. A plan file is an ASCII representation of a Hierarchical Task Network (HTN) that details the actions and sequences to complete a task. The actual syntax of the plan file is an extension of the RETSINA and TMS structure detailed in [11, 2] DECAF reads the plan file and determines which actions can be executed in parallel and which ones must wait to be enabled by the output of a previous action. The plan file that is used for BigAgent is a tree that represents approximately 5000 possible execution paths to complete a task named I ....

Keith S. Decker and Victor R. Lesser. Quantitative modeling of complex computational task environments. In Proceedings of the Eleventh National Conference on Artificial Intelligence, pages 217--224, Washington, July 1993.

....desirable properties such as component reuse (eventually, automated via the planner) and some design time error checking. The chaining of activities can involve traditional looping and if then else constructs. This part of DECAF is an extension of the RETSINA and TAEMS task structure frameworks [34, 11]. Unlike traditional software engineering, each action can also have attached to it a performance profile which is then used and updated internally by DECAF to provide real time local scheduling services. The reuse of common agent behaviors is thus increased because the execution of these ....

....capabilties. Similar to DECAF, JACK will analyze plans and make decisions about sequential or paralle execution, respond apropriately in the event of a failure and decide when sufficient conditions exist to enable an action. Functionally, DECAF is based on RETSINA [30, 9, 12, 34, 33] and TAEMS[11, 31]. However, DECAF, has been restructured to provide a platform for rapid development of agents and as a platform for researching specific areas of agent interaction. DECAF is also written in Java 1 , and makes extensive use of the Java threads capabilities to improve performance each agent ....

[Article contains additional citation context not shown here]

Keith S. Decker and Victor R. Lesser. Quantitative modeling of complex computational task environments. In Proceedings of the Eleventh National Conference on Artificial Intelligence, pages 217--224, Washington, July 1993.

No context found.

K. S. Decker and V. R. Lesser. Quantitative modeling of complex computational task environments. In Proceedings of the Eleventh National ConferenceonArti#cial Intelligence, pages 217#224, Washington, July 1993.

....to this basic model, making it more like the DVMT. For example, it adds the characteristic that each agent has two methods with which to deal with sensed data: a normal VLM and a level hopping VLM (the level hopping VLM produces less quality than the full method but requires less time; see (Decker, Lesser Whitehair 1990, Decker, Garvey, Humphrey Lesser 1993) for this and other approximate methods; a similar technique can be used to model agents who have different capabilities such as processor speed) It also adds a representation of faulty sensors (a subjective level effect) and result sharing, as well as ....

Decker, K. S. & Lesser, V. R. (1993d), Quantitative modeling of complex computational task environments, in `Proceedings of the Eleventh National Conference on Artificial Intelligence', Washington, pp. 217--224.

....desirable properties such as component reuse (eventually, automated via the planner) and some design time error checking. The chaining of activities can involve traditional looping and if then else constructs. This part of DECAF is an extension of the RETSINA and TAEMS task structure frameworks [23, 5]. Unlike traditional software engineering, each action can also have attached to it a performance profile which is then used and updated internally by DECAF to provide real time local scheduling services. The reuse of common agent behaviors is thus increased because the execution of these ....

.... to classical AI black and white goals or worth oriented objective functions over states [16, 22] Currently, each capability is represented as a complete task reduction tree (HTN [7] similar to that in RETSINA [22] with annotations drawn from the T MS task structure description language [5, 21]. The leaves of the tree represent basic agent actions (HTN primitive tasks) One agent can have dozens or even hundreds of actions. The basic actions must be programmed in a precise way just as any program written in C or Java must be. However, the expression of a plan for providing a complete ....

Keith S. Decker and Victor R. Lesser. Quantitative modeling of complex computational task environments. In Proceedings of the Eleventh National Conference on Artificial Intelligence, pages 217--224, Washington, July 1993.

....framework for handling the commitments with uncertainty. Experimental results illustrating the strength of our approach, is provided in Section 5. We conclude with a brief summary in Section 6. 2 Uncertainty in Commitments For the purpose of illustration, our discussion uses the T MS framework [4] for modeling the agent task environment. This does not introduce loss of generality because T MS is domain independent and capable of expressing complex task environments. In terms of reasoning and coordination using the T MS, our discussion will focus around the scheduling framework of ....

Keith S. Decker and Victor R. Lesser. Quantitative modeling of complex computational task environments. In Proceedings of the Eleventh National Conference on Artificial Intelligence, 214-217., 1993.

....among them, then combine their partial solutions into an overall solution, subproblem communication and integration protocols can have an enormous impact on problem solving efficiency. While the differential efficacy of various protocols in different situations has been documented (for example, [4, 3]) there is no general theory of how communication and subproblem integration should be This material is based upon work supported by the National Science Foundation under Grants No. IRI 9321324 and IRI 9523419. Any opinions, finding, and conclusions or recommendations expressed in this ....

K. S. Decker and V. R. Lesser. Quantitative modeling of complex computational task environments. In Proceedings of the Eleventh National Conference on Artificial Intelligence, pages 217--224, Washington, DC, USA, 1993.

....complex, multi level, abstract plans and sophisticated coordination across plans. These tradeoffs seem to be reasonable for an information agent (where they would not be for a higherlevel task agent) Our information agent architecture also extends the T MS task structure representation [4] to allow periodic, repeating actions. 3.1 Control: Planning, Scheduling, and Action Execution The simplified control process for information agents includes steps for planning to achieve local or non local goals, scheduling the actions within these plans, and actually carrying out these actions. ....

....planning process (see Figure 2) takes as input the agent s current set of goals G (including any new, unplanned for goals G n ) and the set of current task structures (plan instances) T . It produces a new set of current task structures. Information agents use simple task structures (compare to [4]) ffl Each individual task T represents an instantiated approach to achieving one or more of the agent s goals G it is a unit of goal directed behavior. Every task has an (optional) deadline. ffl Each task consists of a partially ordered set of basic actions A. These actions are ordered on a ....

Keith S. Decker and Victor R. Lesser. Quantitative modeling of complex computational task environments. In Proceedings of the Eleventh National Conference on Artificial Intelligence, pages 217--224, Washington, July 1993.

....a suitable subset of the coordination mechanisms based on the present problem solving situation, instead of having a fixed subset across all problem instances in an environment. GPGP coordination mechanisms rely on the coordination problem instance being represented in a framework called T MS[4, 5]. The T MS framework (Task Analysis, Environment Modeling, and Simulation) 4, 5] represents coordination problems in a formal, domain independent way. It captures the essence of the coordination problem instances, independent of the details of the domain, in terms of trade offs between multiple ....

....solving situation, instead of having a fixed subset across all problem instances in an environment. GPGP coordination mechanisms rely on the coordination problem instance being represented in a framework called T MS[4, 5] The T MS framework (Task Analysis, Environment Modeling, and Simulation) [4, 5] represents coordination problems in a formal, domain independent way. It captures the essence of the coordination problem instances, independent of the details of the domain, in terms of trade offs between multiple ways of accomplishing a goal, progress towards the achievement of goals, various ....

[Article contains additional citation context not shown here]

Keith S. Decker and Victor R. Lesser. Quantitative modeling of complex computational task environments. In Proceedings of the Eleventh National Conference on Artificial Intelligence, pages 217--224, Washington, July 1993.

....are attempting to formalize and generalize the results from these studies. The results discussed in this paper are important though limited by a single instance experimental methodology. We are developing an abstract task generator capable of capturing all the interactions of KSIs in any domain [10]. We have also worked on quantifying these interactions (called coordination relationships) for distributed and parallel processing[8] Work developing a parallel scheduling theory in this abstract environment has begun [20] with the hopes of providing general results with wider applicability. ....

Keith S. Decker and Victor R. Lesser. Quantitative modeling of complex computational task environments. In Proceedings of the Eleventh National Conference on Artificial Intelligence, Washington, July 1993.

....with a discussion of future work. 2 Proposals and Explanations Problem solving begins when a problem to be solved arrives at the decision maker. In our work, problems are presented to the decision maker as T MS task structures. The form of such task structures is described in more detail in [Decker and Lesser, 1993] . Briefly, a problem episode E consists of a set of independent task groups E = hT 1 ; T 2 ; T n i, each with a hard deadline D(T ) and containing interrelated tasks T 2 T . Within a task group, tasks form a directed acyclic graph through the subtask relationship. The quality or value of ....

Keith S. Decker and Victor R. Lesser. Quantitative modeling of complex computational task environments. In Proceedings of the Eleventh National Conference on Artificial Intelligence, pages 217--224, Washington, July 1993.

....are subtask and dependency. A reasonable parallel schedule can be generated only if the scheduler has taken into account these relationship. For example, if there is a dependency relationship between two tasks, the dependent task cannot be executed before the other is done. Decker and Lesser [7] have identified several kinds of other task relationships which also affect the quality of a schedule. Among them, there are two that are most relevant here: Facilitates Task A facilitates Task B, if execution of Task A will decrease the duration of executing Task B, or increase its quality. ....

....schedule in section 5. Throughout the paper, we will use examples to explain how each algorithm works. Section 7 reviews related work, and section 8 analyzes our approach and offers suggestions for future research. 2 Environment and Assumptions Our scheduler works in a modified T MS environment [7]. The scheduler is given a set of task groups that are represented in T MS task structure, as shown in Fig. 1. The leaves in a task structure are the executable methods. The other non leaf nodes act as quality accumulation functions; they represent tasks that achieve quality from their subtasks ....

Keith S. Decker and Victor R. Lesser. Quantitative modeling of complex computational task environments. In Proceedings of the Eleventh National Conference on Artificial Intelligence, pages 217--224, July 1993.

....with the learning presented here on situation based choice of coordination could have interesting implications for situation specific learning. 3 Task Analysis, Environment Modeling, and Simulation 3. 1 TAEMS The T MS framework (Task Analysis, Environment Modeling, and Simulation) [Decker Lesser, 1993] represents coordination problems in a formal, domain independent way. A T MS model of a task environment specifies what actions are available to agents and how those actions relate to one another and to the performance of the system as a whole. A coordination problem instance is represented as a ....

....have base level quality and duration. A task may have multiple ways to accomplish it, represented by multiple methods, that trade off the time to produce a result for the quality of the result. Besides tasks subtask relationships, there can be other interrelationships between tasks in a task group[Decker Lesser, 1993] . In this paper, we will be dealing with two such interrelationships: ffl facilitates relationship or soft interrelationship: If the results of execution of Task A are available for Task B before it starts executing, then Task B will have increased quality and or decreased duration. ffl ....

Decker, K. S., and Lesser, V. R. 1993. Quantitative modeling of complex computational task environments. In Proceedings of the Eleventh National Conference on Artificial Intelligence, 217--224.

....a suitable subset of the coordination mechanisms based on the present problem solving situation, instead of having a fixed subset across all problem instances in an environment. GPGP coordination mechanisms rely on the coordination problem instance being represented in a framework called T MS(Decker Lesser 1993). The T MS framework (Task Analysis, Environment Modeling, and Simulation) Decker Lesser 1993) represents coordination problems in a formal, domain independent way. It captures the essence of the coordination problem instances, independent of the details of the domain, in terms of trade offs ....

....instead of having a fixed subset across all problem instances in an environment. GPGP coordination mechanisms rely on the coordination problem instance being represented in a framework called T MS(Decker Lesser 1993) The T MS framework (Task Analysis, Environment Modeling, and Simulation) (Decker Lesser 1993) represents coordination problems in a formal, domain independent way. It captures the essence of the coordination problem instances, independent of the details of the domain, in terms of trade offs between multiple ways of accomplishing a goal, progress towards the achievement of goals, various ....

[Article contains additional citation context not shown here]

Decker, K. S., and Lesser, V. R. 1993. Quantitative modeling of complex computational task environments. In Proceedings of the Eleventh National Conference on Artificial Intelligence, 217--224.

....and therefore reduce the time required to achieve a global solution. Also, it may be the case that a solution or partial solution generated by one agent may facilitate (i.e. serve to focus or constrain) the problem solving of another agent and thereby reducing the amount of computation required[10]. The problem solving process of one agent in some way assists another agent in its problem solving, perhaps by making the other agent more certain of its local solution or by restricting the space of potential solutions that must be considered. The end effect is a better or higher quality ....

....is the explicit recognition and exploitation or avoidance of redundancy, leading to increased robustness or decreased resource demands depending on the context and the structure of the domain. We now discuss implemented aspects of FA C that have direct relevance to the CIG task. Decker and Lesser[9, 10, 11] provide detailed studies of quantitative trade offs involved in explicit recognition and exploitation of task interrelationships for use in multi agent coordination. Von Martial s work[26] on coordination in multi agent planning using favors goal interrelationships and temporal interactions is ....

K. Decker and V. R. Lesser, " Quantitative Modeling of Complex Computational Task Environments", in Proceedings of the Eleventh National Conference on Artificial Intelligence, pp 217-224, Washington, 1993.

No context found.

Keith S. Decker and Victor R. Lesser. Quantitative modeling of complex computational task environments. In Proceedings of AAAI-93, pages 217--224, Washington, D.C. Alan Garvey and Victor Lesser. Design-to-time real-time scheduling. IEEE Transactions on Systems, Man and Cybernetics, 23(6):1491--1502, 1993.

....does not necessarily reflect the position or the policy of the Government and no official endorsement should be inferred. must be heeded to find correct solutions (e.g. hard precedence constraints) and soft interactions that can improve (or hinder) performance (e.g. facilitates constraints[Decker and Lesser, 1993] ) An example of a problem to be solved by the design totime scheduling algorithm is given in Figure 1. This representation of a task structure is based on the T MS modeling framework[Decker and Lesser, 1993] In a T MS task structure the leaves of the graph represent executable computations ....

....interactions that can improve (or hinder) performance (e.g. facilitates constraints[Decker and Lesser, 1993] An example of a problem to be solved by the design totime scheduling algorithm is given in Figure 1. This representation of a task structure is based on the T MS modeling framework[Decker and Lesser, 1993] . In a T MS task structure the leaves of the graph represent executable computations (known as methods) and the nonleaf nodes represent tasks that achieve quality as a function of the quality of their subtasks. Each separate graph is known as a task group and represents a single independent ....

Decker, Keith S. and Lesser, Victor R. 1993. Quantitative modeling of complex computational task environments.

....representation for coordination problems and the grammar based generators for modeling domain semantics and experimenting with different aspects of such domains. TAEMS: Task Analysis, Environment Modeling, and Simulation The taems framework (Task Analysis, Environment Modeling, and Simulation) (Decker Lesser 1993; Decker 1995) represents coordination problems in a formal, domain independent way. In the simplest terms, a taems model of a task environment specifies what actions are available to agents and how those actions relate to one another and to the performance of the system as a whole. We have used ....

....for capturing a relevant domain theory for an environment without extensive knowledge engineering ffl the generalized partial global planning approach to constructing environment centered collections of coordination mechanisms. Details about the taems representation and GPGP can be found in (Decker Lesser 1993; 1995; Decker 1995) In particular, this paper also considered a particular problem, functionally organized Distributed Data Processing (DDP) in some detail. We investigated the behavior of three different coordination algorithms for a set of task structures in DDP. We empirically demonstrated ....

Decker, K. S., and Lesser, V. R. 1993. Quantitative modeling of complex computational task environments.

No context found.

K. S. Decker and V. R. Lesser. Quantitative modeling of complex computational task environments. In Proceedings of the 11th National Conference on AI, 1993.

No context found.

K. S. Decker and V. R. Lesser, "Quantitative modeling of complex computational task environments," in Proceedings of the 11th National Conference on Artificial Intelligence, 1993.

No context found.

K. S. Decker and V. R. Lesser, "Quantitative modeling of complex computational task environments," in Proceedings of the 11th National Conference on Artificial Intelligence, 1993.

No context found.

Decker, K.S., and Lesser, V.R.: Quantitative modeling of complex computational task environments. Proceedings of the Eleventh National Conference on Artificial Intelligence, Washington, (1993) 217-224.

No context found.

K. S. Decker and V. R. Lesser. Quantitative modeling of complex computational task environments. In Proceedings of the 11th National Conference on AI, 1993.

No context found.

K. Decker and V. R. Lesser, "Quantitative Modeling of Complex Computational task Environments," Proc. of the AAAI 1993.

No context found.

K. Decker and V. Lesser. Quantitative Modeling of Complex Computational Task Environments, Proceedings of the Eleventh National Conference on Artificial Intelligence, 1993.

No context found.

K. Decker and V. Lesser. Quantitative modeling of complex computational task environments. Proceedings of the Eleventh National Conference on Artificial Intelligence, 1993.

No context found.

Decker, K. & Lesser, V. (1993b), Quantitative Modeling of Complex Computational Task Environments, in `Proceedings, AAAI-93', Washington, DC, pp. 217--224.

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