DMCA
Using Pareto Optimality to Coordinate Distributed Agents (1995)
Venue: | Arti Intelligence for Engineering Design, Analysis and Manufacturing (AIEDAM |
Citations: | 21 - 6 self |
Citations
594 |
gIBIS: A hypertext tool for exploratory policy discussion
- Conklin, Begeman
- 1988
(Show Context)
Citation Context ...f normal engineering discussion. An objective function is simply not an appropriate tool for support of such conflicts. There are decision support tools for providing argumentation in such a situation=-=[1, 15]-=- and for attempting to optimize multiple objective decisions[7]. It is sometimes feasible to at least perform local optimization over weighted objectives[2]. We do not here address all of the problems... |
52 |
Welfare economics and social choice theory.
- Feldman
- 1980
(Show Context)
Citation Context ...uch computer research on support of collaborative engineering concerns global metrics for optimization, decision support, and negotiation, a basic coordination function is support of Pareto optimality=-=[5]-=-. It is frequently difficult to find a global objective function even for problems that otherwise can be easily mapped into integer programming (IP), a general technique for satisfying multiple object... |
39 |
Making compromises among antagonist constraints
- Descottes, Latombe
- 1985
(Show Context)
Citation Context ...viding argumentation in such a situation[1, 15] and for attempting to optimize multiple objective decisions[7]. It is sometimes feasible to at least perform local optimization over weighted objectives=-=[2]-=-. We do not here address all of the problems in these approaches beyond noting that there seems to be no general solution to multiple objective optimization. However, there is a simpler but useful for... |
28 |
Constrained Decision Revision
- Petrie
- 1992
(Show Context)
Citation Context ...ment a principle such as Pareto optimality, one must have a formal notion of an objective, the satisfaction of an objective, and conflicts between objectives. The Redux model of design problem solving=-=[12]-=- provides such formality. Only a subset of this model is needed to perform the bookkeeping duties described above. We call this subset Redux 0 and have implemented it as a coordination service agent[1... |
24 | Reasoned assumptions and Pareto optimality
- Doyle
- 1985
(Show Context)
Citation Context ...ncompleteness. And, in the Redux model interpretation, it is also the mechanism needed to track Pareto optimality. Though Pareto optimality has been previously connected to a kind of truth maintenance=-=[4]-=-, this mapping between dependency-directed backtracking and Pareto 19 optimality is novel. 5 The Next-Link Framework How is such a model to be practically implemented for distributed engineering? As w... |
22 | An Agent-Based Approach to Concurrent Cable Harness Design,
- Park, Cutkosky, et al.
- 1994
(Show Context)
Citation Context ...mechanism already existed in the interface. 10 See WWW URL http://cdr.stanford.edu/html/NextLink/NextLink.html. In addition, the general problem domain has been studied in a project called First-Link =-=[9]-=-. Here we briefly remark that the original research has spawned four separate projects. There is a commercial version of the First-Link system being used at Lockheed Missile Systems, which supported t... |
22 | Design Space Navigation as a Collaborative Aid
- Petrie, Cutkosky, et al.
- 1994
(Show Context)
Citation Context ...erform the bookkeeping duties described above. We call this subset Redux 0 and have implemented it as a coordination service agent[13] that performs a number of useful functions for distributed design=-=[14]-=-. In this section, we describe this model and show how in particular it supports the tracking of Pareto optimality. To simplify matters, we do not distinguish further the complete problem-solving Redu... |
22 |
Representing and maintaining process knowledge for large-scale systems development
- Ramaesh, Dhar
- 1994
(Show Context)
Citation Context ...f normal engineering discussion. An objective function is simply not an appropriate tool for support of such conflicts. There are decision support tools for providing argumentation in such a situation=-=[1, 15]-=- and for attempting to optimize multiple objective decisions[7]. It is sometimes feasible to at least perform local optimization over weighted objectives[2]. We do not here address all of the problems... |
21 | The Redux’ Server
- Petrie
- 1993
(Show Context)
Citation Context ...2] provides such formality. Only a subset of this model is needed to perform the bookkeeping duties described above. We call this subset Redux 0 and have implemented it as a coordination service agent=-=[13]-=- that performs a number of useful functions for distributed design[14]. In this section, we describe this model and show how in particular it supports the tracking of Pareto optimality. To simplify ma... |
12 |
Context Maintenance.
- Petrie
- 1991
(Show Context)
Citation Context ...ions, as opposed to, say, chronological backtracking. However, it implies a search mechanism that ensures that one does not jump back to a search state previously visited nor omit a new possible state=-=[11]-=-. To elaborate briefly on the rationale to be constructed for a rejected decision, given a constraint C that is violated by the conjunction of a set of assignments fA 1 : : : A n g, there is a corresp... |
9 |
A Multiple Objective Linear Programming Decision Support System
- Korhonen, Wallenius
- 1990
(Show Context)
Citation Context ...t an appropriate tool for support of such conflicts. There are decision support tools for providing argumentation in such a situation[1, 15] and for attempting to optimize multiple objective decisions=-=[7]-=-. It is sometimes feasible to at least perform local optimization over weighted objectives[2]. We do not here address all of the problems in these approaches beyond noting that there seems to be no ge... |
6 |
SHADE: A Medium for Sharing Design Knowledge Among Engineering Tools
- McGuire, Pelavin, et al.
- 1993
(Show Context)
Citation Context ...f the current Next-Link system is shown in Figure 13. Each agent in the architecture has a basic API for communication. In the current implementation, this is KAPI: the KQML API from the SHADE project=-=[6]-=-. This basic 22 DECISION TASK COMPLETION Next-Link Architecture Redux' API PS Redux' API DM Redux' API CE Redux' API FSM Redux' API EE CM: Constraint Manager CONSTRAINT REDUX' Design Dependency Mainte... |
5 | Comparative analysis of design rationale representations
- Lee, Lai
(Show Context)
Citation Context ... We will briefly discuss the first two but emphasize the last in this paper. To this end, we now suppose that the restricted zone 3 The system described accommodates the usual sort of design rationale=-=[8]-=- but provides additional important information about the history of the design process. 7 M3 M2 M1 M6 M4 M5 High Temperature Zone Con1 Con2 Con3 Con4 Con7 Con8 Con9 Con6 X Figure 6: Conflict: P1V5 & P... |
4 | Modeling of Collaborative Design Processes for Agent-Assisted Product Design - Park - 1995 |
2 |
An Experiment
- Dhar, Raganathan
- 1990
(Show Context)
Citation Context ...e to market, weight, or various features? The difficulty of generating an "objective" function is increased by the fact that even small numeric changes in weights can generate very different=-= solutions[3]-=-. When problem solving is distributed over multiple agents with expertise in different domains, the difficulty is exacerbated. For example, an electronics engineer may want to use a position sensor th... |
1 |
8 Author Biographies Charles Petrie is a Senior Research Associate at the Center for Design Research at Stanford University. Prior to this, Dr. Petrie was a founding member of the AI lab at the Microelectronics and Computer Technology Corporation (MCC). H
- Wilkens
- 1988
(Show Context)
Citation Context ...set of users will not generate the same conflict over and over, perhaps in a cycle that is not easily predicted. This guarantee is unique to Redux: it is not true of systems such as that 16 of [2] or =-=[16]-=- or any other that supports this kind of revision capability. 4.2 The AI Substrate Redux represents work from artificial intelligence in two major ways. One is the model ontology. It was developed out... |