E. Durfee and V. Lesser. Predictabilityvs.responsiveness: Coordinating problem solvers in dynamic domains. In ########### ## ### ####### ######## ########## ## ######### ############, pages 66-71, St. Paul, Minnesota, August 1988.  Home/Search   Document Not in Database   Summary   Related Articles   Check

....our existing optimizations, in large numbers these actions can still consume a signi cant portion of the agents operating time. To compensate for this, we scale the agents time granularityby some xed amount. This theoretically trades o prediction and scheduling accuracy for responsiveness [3], but in practice a suitably chosen value has few drawbacks, because the agent is e ectively already operating at a lower granularity due to the real time missed between agent activity cycles. Using this scheme, if wesay that every agenttick corresponds to 20 milliseconds, the above action would ....

E. Durfee and V. Lesser. Predictabilityvs.responsiveness: Coordinating problem solvers in dynamic domains. In ########### ## ### ####### ######## ########## ## ######### ############, pages 66-71, St. Paul, Minnesota, August 1988.

....model for designing the distributed planning mechanism, however we use a physics oriented approach for cooperative MAS, not for competitive agents. A large body of DAI research studies coordination among agents for distributed problem solving (for example, 2] PGP [5] GPGP [1] 6] 18] In [4], Durfee and Lesser study the Partial Global Planning (PGP) approach to coordination by implementing it in the Distributed Vehicle Monitoring Testbed (DVMT) The DVMT is a network of vehicle monitoring nodes. Each node has a planner that plans incrementally. Nodes do not communicate their detailed ....

E. Durfee and V. Lesser. Predictability vs. responsiveness: Coordinating problem solvers in dynamic domains. In Proceedings of the Seventh National Conference on Artificial Intelligence, pages 66--71, St. Paul, MN, August 1988.

.... The PAS is applicable for varying, non uniform distributions of masses and locations of particles (and, respectively, agents and tasks) 8 Related work A large body of DAI research studies coordination among agents for distributed problem solving (for example [13] PGP [16] GPGP [9,17,63] In [12], Durfee and Lesser study their Partial Global Planning (PGP) approach to coordination by implementing it in the Distributed Vehicle Monitoring Testbed (DVMT) The DVMT is a network of vehicle monitoring nodes. Each node has a planner that plans incrementally. Nodes do not communicate their ....

E. Durfee and V. Lesser. Predictability vs. responsiveness: Coordinating problem solvers in dynamic domains. In Proceedings of the Seventh National Conference on Artificial Intelligence, pages 66--71, St. Paul, MN, August 1988.

....model for designing the distributed planning mechanism, however we use a physics oriented approach for cooperative MAS, not for competitive agents. A large body of DAI research studies coordination among agents for distributed problem solving (for example, 2] PGP [5] GPGP [1] 6] 18] In [4], Durfee and Lesser study the Partial Global Planning (PGP) approach to coordination by implementing it in the Distributed Vehicle Monitoring Testbed (DVMT) The DVMT is a network of vehicle monitoring nodes. Each node has a planner that plans incrementally. Nodes do not communicate their detailed ....

E. Durfee and V. Lesser. Predictability vs. responsiveness: Coordinating problem solvers in dynamic domains. In Proceedings of the Seventh National Conference on Artificial Intelligence, pages 66--71, St. Paul, MN, August 1988.

.... effects of their actions) When network, data, and problem solving characteristics change and communication channels have delay and limited capacity, nodes can locally respond to new situations, still cooperating but with potentially less effectiveness because they have somewhat inconsistent PGPs [Durfee and Lesser, 1988a] The PGP framework does not deal with conflicts in physical resources. The original PGP algorithm orders intermediate goals according to their cost as computed from the cross product of a vector of seven computable factors (described below) and global cooperation parameters that give a weight ....

.... dynamically (as an agent reacts to the problem being solved) The node plan structure can be modified to hold ranges as well as a best current estimate for a value, but it is also likely that agents will have to reason and perhaps negotiate about predictability versus reliability issues as well [Durfee and Lesser, 1988a] The node plan structure could also be expanded with contingency plans for routine expectation failures [Dean, 1987] to allow for predictability in the face of a changing environment. Agents can also make commitments to certain task characteristics, and add explicit slack to schedules[Decker ....

E. Durfee and V. Lesser. Predictability vs. responsiveness: Coordinating problem solvers in dynamic domains. In Proceedings of the Seventh National Conference on Artificial Intelligence, pages 66--71, St. Paul, Minnesota, August 1988.

....events are individually quite fast, combined in large numbers they can consume a signi cant portion of the agents operating time. To compensate for this, we scale the agents time granularity by some xed amount. This theoretically trades o prediction and scheduling accuracy for responsiveness [6, 7], but in practice a suitably chosen value has few drawbacks, because the agent is e ectively already operating at a lower granularity due to the real time missed between agent activity cycles. Using this scheme, if we say that every agent tick corresponds to 20 milliseconds, the above action would ....

E. Durfee and V. Lesser. Predictability vs. responsiveness: Coordinating problem solvers in dynamic domains. In Proceedings of the Seventh National Conference on Arti cial Intelligence, pages 66-71, St. Paul, Minnesota, August 1988.

....when it is, it can adjust the frequency of monitoring to optimize utility. This work has significant overlap with the foundations of my work. However, it is in a multi agent non anytime setup with interacting agents which makes the decision making process more complex. The partial global planning [8] approach is a flexible framework for coordination where nodes can balance their needs for predictability and responsiveness differently for different situations. In this framework, nodes exchange information about their tentative local plans and develop partial global plans(PGPs) to represent the ....

E. Durfee and V. Lesser. Predictability vs. responsiveness: Coordinating problem solvers in dynamic domains. pages 66--71, St. Paul, Minnesota, August 1988.

....not provide a deep analysis of when and why they are appropriate. We believe that the right way to think about coordination is through abstractions of these coordination relationships, defined in a domain independent way. These relationships need to be quantified, not just identified. Earlier work [8, 10] has shown that a weakly qualitative approach to answering questions about coordination can lead to unsatisfying answers that different coordination algorithms (organizations, communication patterns, etc. are better or worse depending on the situation. These experiences and the large parameter ....

....a commitment to impact many other unknown agents, a break in the commitment of least benefit is not liable to have this effect due to the slack and resiliency of each successive local scheduler. Slack has long been viewed as a coordination behavior and as such should be an organizational parameter [8]. This treatment does not give full justice to the analysis of this particular problem, to which we will have to devote a future analysis of its own. 9 4 Hypotheses and Experiments As we flesh out our conceptual model to a set of analytic components by making various assumptions, we develop a ....

[Article contains additional citation context not shown here]

E. Durfee and V. Lesser. Predictability vs. responsiveness: Coordinating problem solvers in dynamic domains. In Proceedings of the Seventh National Conference on Artificial Intelligence, pages 66--71, August 1988.

....all environments. In human organizations, environmental factors such as dynamism and task uncertainty have a strong effect on what coordinated actions are and how organizationally acceptable outcomes arise[18, 9, 33] These effects have been observed in purely computational organizations as well[8, 7, 6, 23, 26]. Achieving effective coordination in a multi agent system (MAS) is a difficult problem for a number of reasons. An agent s local control decisions about what activity to do next or what information to communicate and to whom or what information to ask others may be inappropriate or suboptimal due ....

....strategies to acquire non local views of task structures may not be worthwhile for all problem solving situations. In certain situations, coordination protocols that permit some level of non coherent activity and avoid the additional overhead for coordination may lead to better performance[7, 6, 23, 34]. For example, when the agents are under severe time pressure and the load of the activities at the agents is high, sophisticated agent coordination strategies do not generally payoff. Agents may not have the time to benefit from the increased awareness they derive through coordination in such ....

E. Durfee and V. Lesser. Predictability vs. responsiveness: Coordinating problem solvers in dynamic domains. In Proceedings of the Seventh National Conference on Artificial Intelligence, pages 66--71, St. Paul, Minnesota, August 1988.

.... effects of their actions) When network, data, and problem solving characteristics change and communication channels have delay and limited capacity, nodes can locally respond to new situations, still cooperating but with potentially less effectiveness because they have somewhat inconsistent PGPs [13]. The PGP framework does not deal with conflicts in physical resources. 3 Issues in Extending the PGP Mechanisms The global coherence problems we would like to address occur in many systems other than the DVMT, such as the Pilot s Associate (PA) system [36] where situations occur that cause ....

.... dynamically (as an agent reacts to the problem being solved) The node plan structure can be modified to hold ranges as well as a best current estimate for a value, but it is also likely that agents will have to reason and perhaps negotiate about predictability versus reliability issues as well [13]. The node plan structure could also be expanded with contingency plans for routine expectation failures [7] to allow for predictability in the face of a changing environment. Agents can also make commitments to certain goal characteristics, and add explicit slack to schedules[11] 3.3 Real time ....

E. Durfee and V. Lesser. Predictability vs. responsiveness: Coordinating problem solvers in dynamic domains. In Proceedings of the Seventh National Conference on Artificial Intelligence, pages 66--71, August 1988.

....However, in situations where rescheduling is not appealing, the scheduler can create less efficient schedules that are fault tolerant by making conservative probabilistic assumptions about quality, cost, duration and uncertainty. This relates somewhat to previous work done by Durfee and Lesser [8] and in [9] in which schedules are made loose by increasing duration expectations when building schedules, effectively creating a slack time buffer between each action. Our model is much stronger in that we change expectations based on probabilities rather than using magic numbers, and we do so ....

E. Durfee and V. Lesser. Predictability vs. responsiveness: Coordinating problem solvers in dynamic domains. In Proceedings of the Seventh National Conference on Artificial Intelligence, pages 66--71, St. Paul, Minnesota, August 1988.

....metalevel processing, it may not be worthwhile to acquire a complete view of other agents activities. Sophisticated coordination strategies are not cost effective in all problem solving situations, and permitting some level of non coherent activity to occur may be the optimal coordination strategy[4]. In this case, local problem solving is done more efficiently where there is no additional overhead for coordination. Generalized Partial Global Planning (GPGP) has recognized this need for creating tailored coordination strategies in response to the characteristics of a particular task ....

E. Durfee and V. Lesser. Predictability vs. responsiveness: Coordinating problem solvers in d ynamic domains. In Proceedings of the Seventh National Conference on Artificial Intelligence, pages 66--71, St. Paul, Minnesota, August 1988.

.... such as dynamism and task uncertainty have a strong effect on what coordinated actions are and how organizationally acceptable outcomes arise[Lawrence and Lorsch, 1967, Galbraith, 1977, Stinchcombe, 1990] These effects have been observed in purely computational organizations as well[Fox, 1981, Durfee and Lesser, 1988, Decker and Lesser, 1995, Nagendra Prasad et al. 1996a, Nagendra Prasad et al. 1996b] Achieving effective coordination in a multi agent system (MAS) is a difficult problem for a number of reasons. An agent s local control decisions about what activity to do next or what information to ....

....strategies to acquire non local views of task structures may not be worthwhile for all problem solving situations. In certain situations, coordination protocols that permit some level of non coherent activity and avoid the additional overhead for coordination may lead to better performance[Durfee and Lesser, 1988, Decker and Lesser, 1995, Nagendra Prasad et al. 1996a, Sugawara and Lesser, 1993] For example, when the agents are 71 under severe time pressure and the load of the activities at the agents is high, sophisticated agent coordination strategies do not generally payoff. Agents may not have ....

Durfee, E. and Lesser, V. Predictability vs. responsiveness: Coordinating problem solvers in dynamic domains. In Proceedings of the Seventh National Conference on Artificial Intelligence, pages 66--71, St. Paul, Minnesota, August 1988.

....National Science Foundation under Grant Nos. IRI 9523419 and EEC 9209623. The content of this paper does not necessarily reflect the position or the policy of the Government, and no official endorsement should be inferred. non coherent activity to occur may be the optimal coordination strategy[7]. In this case, local problem solving is done more efficiently where there is no additional overhead for coordination. Nagendra Prasad et al. 11] developed a tool for representing interesting, complex coordination problems and instantiating environment specific coordination mechanisms for ....

E. Durfee and V. Lesser. Predictability vs. responsiveness: Coordinating problem solvers in dynamic domains. In Proceedings of the Seventh National Conference on Artificial Intelligence, pages 66--71, St. Paul, Minnesota, August 1988.

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